Elon Musk and OpenAI will go to trial this week in a case that should be filled with drama and juicy revelations.

You’re probably going to hear a lot about personality conflict between Musk and OpenAI CEO Sam Altman. But here’s what’s at stake in the case, and what the outcome might be:

This is a legal battle over whether OpenAI lawfully evolved from a nonprofit, breaching an agreement to operate as a charity that the founders — including Musk, who invested $38 million — made at its inception.

Musk’s lawsuit hinges on two claims: Breach of charitable trust and unjust enrichment. OpenAI has denied the allegations, arguing that its structure had to evolve to recruit talent and build new AI systems. It also countersued Musk and xAI, claiming they interfered in OpenAI’s relationships with investors, customers, and employees.

The stakes are high. Musk is asking the court for OpenAI and Microsoft to give up as much as $134 billion, which he says should be directed to OpenAI’s charitable mission through a restructuring that would force it to operate entirely as a nonprofit.

If the case reaches the remedy phase, the judge would decide how much each company would pay. Musk is also asking the court to remove Altman and OpenAI President Greg Brockman from the company.

The court could hear testimony from a list of high-profile tech execs that includes Musk, Altman, Brockman, Microsoft CEO Satya Nadella, former OpenAI CTO Mira Murati and OpenAI co-founder Ilya Sutskever. Others on the list include ex-OpenAI board members Tasha McCauley and Helen Toner.

Timeline & context for the case:

• The liability phase will be heard by an advisory jury, which would offer a nonbinding verdict for the judge to consider.

• The judge plans to give Musk and OpenAI each 22 hours to present their case in the liability phase, with Microsoft getting 5 hours.

• If Musk gets past the liability phase, the remedies phase is expected to begin May 18, which would have new time allotments for each party.

Some are placing bets on whether Elon will win his case. As of Sunday night, around $390,000 in bets have been placed on Kalshi since mid-January, with the platform estimating his chances at 49.9%.

Pre-trial filings are full of colorful claims in documents from all parties, including emails and other rhetoric. Here’s two:

“Obviously we’d comply with/aggressively support all regulation.” - Sam Altman (May 2015) in a May 2015 email to Elon Musk.

“Because we don’t have any financial obligations, we can focus on the maximal positive human impact and disseminating AI technology as broadly as possible.” — Sam Altman to Elon Musk email (December 2015)

A MESSAGE FROM OUR SPONSOR

AI, Power, and the Future of Democracy

Join Orchestra and Big Technology in San Francisco on May 12 for a conversation with Alex Kantrowitz, Orchestra CEO Jonathan Rosen, and Orchestra Partner David Plouffe, exploring the forces reshaping the relationship between technology and democracy.

Save your spot.

The Intelligence Report

• OpenAI’s exclusivity agreement with Microsoft is over, making room for Amazon and Google to sell its models via their cloud platforms.

• On Sunday night, less than 24 hours before OpenAI heads to court, OpenAI published a new blog post from Sam Altman titled “Our Principles.”

• Meta and Amazon signed a multibillion dollar deal for AWS to power the social giant’s agentic AI efforts using Amazon’s Graviton CPU chips.

• Google debuted a number of updates at Cloud Next 2026, including two new chips, a new Gemini Enterprise Agent Platform, a new Gemini Enterprise App, and an agentic data cloud.

• OpenAI debuted GPT-5.5 (aka “spud”), which you can hear more about in Big Technology’s special edition podcast episode with Greg Brockman.

• Meta reportedly plans to start installing tracking software on employee computers to monitor mouse movements and keystrokes for AI training data.

• Google announced plans to invest up to $40 billion in Anthropic just days after Amazon announced it would invest another $5 billion now and up to $20 billion more later on.

• White House pushed out Collin Burns, a former Anthropic safety researcher, just four days into his new role on an AI safety team with the Commerce Department.

• Nvidia-baked VAST Data raised $1 billion and is now worth $30 billion, a sign of how fast AI infrastructure companies are growing today.

• Alex appeared on CNBC to discuss the Musk vs. OpenAI case this morning.

• Listen to this great report about the case from NPR’s John Ruwitch (Alex is quoted).

• It’s once again earnings week. Alphabet, Amazon, Microsoft and Meta are all scheduled to report results April 29. Apple, Samsung, Reddit and Roblox will all go a day later.

Join Big Technology’s Private Discord Server!

Where we’ll talk about this story, the latest in AI, the week’s podcast, and plenty more. You can sign up via the link below:

Read more

OpenAI released its long-awaited GPT-5.5, or “Spud,” model on Thursday. The model is “a beginning point,” OpenAI President Greg Brockman told me shortly after release, setting up a series built to take action on your behalf and accomplish your goals. GPT-5.5, Brockman said, is adept at coding and computer control, able to act without arduous amounts of instruction.

“In many ways, it is a step towards a new way of getting work done with a computer,” Brockman said. “Being very proactive and really being able to solve problems end to end with a little instruction.”

In an ‘emergency’ episode of Big Technology Podcast, I spoke with Brockman about GPT-5.5’s capabilities, whether model commoditization is coming, how Spud plays in the ongoing cybersecurity conversation, and what he means when he says we’re heading toward a “Compute-powered economy.”

You can read our conversation, edited lightly for length and clarity, below. Or listen on Apple Podcasts, Spotify, or your podcast app of choice.

Alex Kantrowitz: Welcome Greg. First off, can you confirm GPT-5.5 is indeed ‘Spud?’

Greg Brockman: Yes

Okay, so what is GPT-5.5?

I think in many ways, it is a step towards a new way of getting work done with a computer. It’s a new class of intelligence. It’s extremely useful at things like programming, and all the different aspects of debugging and solving very hard and gnarly problems, just being very proactive and really being able to solve problems end to end with a little instruction.

But the thing that’s most remarkable is not necessarily the fact that it got better at coding. That, I think, is what everyone kind of expects.. But the fact that it’s now really crossed the threshold of usefulness for general kinds of applications. And so it’s much better at creating slides, spreadsheets, much better at computer use, using your browser, being able to kind of click through applications that are otherwise hard to have an AI operate. And so I think that we’re really seeing the emergence of this new way of using a computer, and it starts with this kind of intelligence at the core.

When we spoke last you mentioned that this was effectively the culmination of a two-year research process. This was planned two-years ago?

Yes, we do have very long horizons for how we plan now. One note is that we stack together many research ideas and bets on a variety of timescales. And so the way to think about it is that we are making constant progress across every single part of the stack.

And so what GPT-5.5 represents is not an endpoint in many ways. It’s a beginning point. It’s really a step towards the kinds of models that we see coming over even just upcoming months. And I think that you should expect that we are going to have even larger improvements in the capability across a wide variety of these aspects of what the model can do, and that’s something that I think will be very exciting.

Can you share specifically what those aspects are? If this is the beginning, what is it the beginning of?

Think about the models as the brain. You can think about the systems and the harnesses like Codex and the applications, like the Super App, as almost the body around it to make it into a useful AI. And that’s really what’s happening, is a shift from language models being the thing that is produced by labs like ourselves to an AI that’s useful, that’s an assistant that’s out there trying to solve your goal, that’s really operating according to your instruction.

And you can see now, Codex is becoming this app that’s not just for the coders. It’s really for anyone using a computer. And it’s not perfect. There are still some tasks where that it should be able to do it and it doesn’t quite get it right. Sometimes the personality isn’t quite what you wanted. It’s extremely powerful and out there doing a lot of really amazing things, but the way it communicates back to you — you have to still spend some time really trying to read through, okay, exactly how did it solve this problem?

And so these aspects, we know exactly how to make them much better. And we’ve already had a pretty remarkable improvement from 5.4 to 5.5. I think we’re going to have even more remarkable improvements across every single aspect of what makes these models useful.

Internally, we think a lot about the end application. That is one thing that changed for us over the past 12, 18 months, something like that, is that we used to really just be focused on — Let’s improve on the benchmarks. Let’s make these models more cerebrally capable. But we now are really focused on — Let’s bring them to real-world application. Let’s think about finance, sales, marketing, every single function that someone uses a computer. How can we help with their computer work? How can we make the model have not just theoretical capability to help but is experienced? Those kinds of tasks, that’s been able to see what good looks like.

And I think that the place we’re going is one where you as a person doing work, that you are the overseer, you are the CEO of almost this autonomous corporation, or, this fleet of agents, perhaps, is more, is the way to say it, and that they are operating according to your goals. Now you are still accountable You’re still in the driver’s seat. You’re still the person who thinks about, well, is this what I wanted, was this work up to standard, but that the details of exactly what buttons were clicked and exactly the code that was written, or exactly how the formula and the spreadsheet works, that you can abstract yourself from those if they’re not important to the evaluation of whether or not something was what you wanted. And so I think it’s like increasing leverage for every worker.

If this is a culmination of two-years of work, is it right to think of it as the moment reinforcement learning training methods overtook pretraining in importance?

I would say it a little differently. There’s many steps to the pipeline. There’s pre-training, mid-training, reinforcement learning, there’s data collection, there’s a lot of these different things that all come together to produce the end result. And the way in which is connected to the world, that’s also very key to making it useful.

The thing that I’m really saying is, we have been investing on every single one of these. We have a team that’s not just about individuals working on these pieces, but a team that really comes together and looks across the whole stack to say, how do we make this more useful for real-world applications?

And so it’s not really any one thing that we do. If you’re building a car— it’s not just about, do you have a better engine. You can build a great engine, but if the rest of the car is not up to the quality level of the engine, it’s not going to matter. And so I think that is the real innovation. It’s really the end-to-end, co-design and all coming together in a repeatable fashion to make these models better and better for our users.

You’ve said the model more intuitively knows what you want. How do you get it to know that? And does that mean prompt engineering is dead?

Number one, I think it really comes down to when we say there’s a new class of capability, a new class of intelligence, that’s really what we mean.. The models are becoming much more intuitive to use because they have deeper understanding of what it is you’re asking of them. They really look at the context, try to understand and puzzle out what am I being asked to do.

And it really makes you realize, to the second part is prompt engineering dead, which I think that prompt engineering, in some ways, may be even more vibrant than before, but you spend so much time now, trying to explain to your computer what you even want. You try to pack in this context and be well, here’s what’s going on, here’s the situation, here’s the thing I want from you. And you’re like — Why do I have to explain this to my computer? Right? The whole thing is the computer should be doing the work to help me. I don’t want to have to be, breaking down the task, trying to explain to it step-by-step how to do things. I want to point it in a direction, and I want it to be able to take care of the details and to get me the result again in a way that I can observe and provide feedback along the way. But I want it to be the driver of the of those like low level execution.

And so I think that in some ways, where prompt engineering is going to go is — it’s going to be about you can get so much more out of these models with so much less effort, but with the same amount of effort, you still have a multiplier.

Okay, let me briefly speak with you about the economics of building a model like this. There’s been this pattern where these big, massive models getting distilled by open source model makers, and then open source is just a couple months behind the leading foundational models. How is this defensible in the long term?

It’s not as simple as you can take the output to these models and distill and you have exactly the model the same capability, it’s just smaller and can run fast. If that were the case, we would just do that, and then we would also have a model that would be much more easy to serve in many ways. And of course, there’s a lot of art behind distillation.

Now the at the deployment side, we think a lot about safeguards. We think a lot about mitigations, and we do that for many, different aspects of how these models could be misused in real situations. And that’s something that we have been investing in for many years, and we think about that across areas like cyber, or thinking about that areas like bio, that we have a long standing effort that you can see in our preparedness framework, which is public, about how we approach these kinds of uses of the model, and how we try to make maximize the benefits, mitigate the risks.

And so I think it’s a real motion that every piece of what we do needs to connect to the question of, how do we continue to make progress, but also, how do we make these models broadly available? Because that’s something that we really believe in, that we believe this technology empowers people, and that we want it to benefit people and lift everyone up.

The pricing on this model is, double the last model, GPT-5.4. And so, from an economics or the business standpoint, if open source can deliver not as good performance, but almost as good, and do it cheaper, how do you handle that threat?

First of all, if you look at our history, which really is not driven by anything in competition. It’s just like our own progress and desire. We have dropped prices on the same level of intelligence year over year, sometimes by literally a factor of 100. It’s like at least an order of magnitude year over year, sometimes literally 100.

The thing that keeps happening, it’s real Jevons paradox, where it’s you lower the cost of something and way more activity happens And I think that what we keep seeing is that there are returns to intelligence that for the kinds of tasks that these models are now capable of doing, that a little bit more intelligence goes a long way.

In some ways, you can almost look at it as oh, there’s just an incremental improvement in intelligence, but I think there’s going to be a massive improvement in terms of what people use it for. And by the way, I think that incremental is very much an understatement for this model relative to five, it’s a point one improvement, in some ways, but I think that really undersells the magic that we see within this model, and that our early testers have really seen in their practical work.

So if people see these numbers and they say — there’s IPO pressure on, OpenAI, and therefore, we’ve been getting a great deal on intelligence, and the free ride is over. You would argue against that?

We have a very simple business in some ways. We rent build by compute, and we resell it with some positive margin, and as long as it’s, positive operating margin, and as long as there’s scalable demand for intelligence, which I think is true, as long as there’s problems to solve, then we can scale that compute all day.

That’s the main directive that I that I ask of the team, is just think about we need to add value onto. Up at the raw compute and make sure that we are at positive operating margin on it, and that is something where it’s not even about the different competition in the marketplace. It’s just a question of, can you have compute that gets turned into intelligence, and that’s just how that it does that at a slightly improved value coming out relative to the cost going in.

And I think that is something where, again, we’re always trying to make more efficient models, but then we just want more of them, and then we want the more intelligent models. And regardless of where they’re coming from, it’s all the same compute that’s going in. And so I think that it’s a great like competition in this marketplace has been great for innovation, but I think that it’s something where it’s driving more usage and more overall spend in the ecosystem, and you can see that in the revenue numbers of us and others in this industry.

Regarding cybersecurity, there are two very different approaches between OpenAI and Anthropic: Anthropic’s latest, massive model, Mythos, is not released to the public. This one, Spud, or GPT-5.5 is released to the public. Is there a chance that releasing this powerful model could lead to some major cyber attacks?

The thing to understand is that we have been investing in cyber safeguards and cybersecurity as a part of our preparedness framework for years. This is something we have invested in far ahead of having the kinds of capabilities we see coming.

And so we have been taking a very deliberate step-by-step approach. You can see, even just over the past couple of weeks, where we’ve expanded our trusted access for cyber program. And, in general, we believe in ecosystem resilience. We think that you do want to go step-by-step, that these models are getting continuously better. We have line of sights and even more capable ones, and that you want to be able to put these models in the hands of defenders to make sure that you’re able to protect critical infrastructure.

And we believe in that resilience of as you can bring these models into people’s hands, that then they’re able to explore in ways that you would not be able to without that access. And so you want this graduated approach and to make sure that you are moving down that pipeline as you can bring in additional safeguards in order to make sure that you can maximize the benefits and mitigate the risks.

And so we’ve really taken a deliberate approach. I think our team has been working incredibly hard to think through the cyber implications of this model. We also believe in iterative deployment. That’s part of this really bringing the models as they continuously get better. And we believe in democratic access. And we believe that ultimately, the goal of creating this technology is to empower people to ensure that it does benefit all of humanity. And so we are constantly trying to solve for how do we safely and responsibly bring this technology to bear in the world in a broad way.

Let me talk through the other case and then get your response. The other case would be, there are you can’t account for everything, and there are clearly going to be some vulnerabilities that will only be found by people or entities deploying this and looking for them. So maybe it makes sense to start with a trusted group of testers before you deploy it, what do you think?

Well, I believe the correct answer here is subtle, and I think it is rooted in the technical specifics of what you have in front of you and many factors. You need to think about, how are the models progressing not just your own capabilities, but others in the ecosystem. You need to think about, what benefit do you get from having a small group that has access and are they able to have high leverage by being able to find and produce patches? But then, how do you coordinate the disclosure of those across an industry?

And so there’s a lot of factors that go into it. I think that the true answer is if either extreme is not quite there are tools that can be applied to. Specific situation, and I think that this is not the first time we’ve had to think about this problem. It’s not the last time we will have to think about it.

But one thing to note is that we have had our model in the hands of defenders for some time, that we’ve been building up our trusted access program, that the model that we’re releasing is not cyber permissive that it has a number of safeguards built into it, and that you can then have a gap between what you’re privately sharing, testing those kinds of things.

And so I think my short answer is, it’s there’s definitely these different schools of thought in terms of values — of is the value that you want to get these models into people’s hands and empower them, or is the value that you want them to be centralized and controlled, and that you don’t want them in people’s hands. That is something that is a maybe underlying tension in some of these debates. But I think that the tactics that those almost flow from the details, and that they can be informed by these values, but either extreme, reflexively, I don’t think will yield the best outcome for the world.

Back to agents. These agents work the best if you let them have a high degree of autonomy. But that’s risky. What is the proper amount of trust to put into agents now?

So I think that now, agents tend to be quite reliable, and even things like prompt injections, I think that there’s still holes there, but that we’re patching them, and that the models are becoming much more resilient.

But I also think that the flip side is that — just like if you have employees, if you have a team of five employees, they’re all trustworthy, fine, but if you have 500,000 of the same employees, that there’s the law of large numbers that you start to worry about, okay, how do I have good governance and oversight.

So this is something where as we’re investing in these capabilities and making the Super App more accessible, not just to coders, but to any person doing work with a computer, we’re also investing in governance and oversight. And you can see this very concretely in workspace agents, which we released recently, that’s within your enterprise, you can now define agents so you get a hosted Codex harness in the cloud. You can hook up tools. You can hook it up to your Slack, and it’s doing work. It’s awesome. A lot of people use it. It’s been very cool to see how viral it goes within an organization. When you see someone else’s agent, you’re like wait, I can build one of these too, and you can just fork it and do your own thing, and then that’s an opportunity to have great governance and that’s baked into the product.

So I think that the short answer is you want to ramp the responsibility entrusted with the agent, and the diversity of things that agents are doing together, with security, safety, observability, oversight, and if you’re not doing those hand in hand, then I think that that’s a little bit out of balance. And I think it’s important to think about both sides.

Go ahead, but be careful?

And really lean in. I think it’s as you scale, you can prototype. That starts to bring in — Do you still have the ability to oversee what’s going on? So you need to make sure at each step. Do you feel like you’re calibrated, you understand what your teams are up to?

Greg, let’s end with this. You’ve called this a “compute powered economy.” What does that mean?

Well, I think we are heading to a world where the more compute is poured into a problem, the faster that problem will be solved, and that the ceiling of problem that can be solved depends on how much compute is available. And you think about things like drug discovery — being able to solve complex diseases like Alzheimer’s is outside of humanity’s reach. Right now. But imagine a world where you can take a gigawatt data center and have it just think about how to solve Alzheimer’s for a month, for a year, however long it takes.

And it may not be literally just cerebrally solving this problem, you may have to consult with world experts. Maybe it has to suggest experiments that get run in a wet lab. But if you can solve such a problem, that would be such a transformatively positive thing for humanity, and I think we were heading to a world where that is how important problems get solved, and that is how tasks in your daily life can also be solved, whether it’s having an agent that knows you, that has your personal context, that is trustworthy, that you can ask for advice on health, and you get back trustworthy information, and that’s just a thing. That’s a smartphone that’s in your pocket You can just talk to and it’ll be out there doing things and proactively knows, what are your goals? What are your interests? And how it can help you.

Compute is going to be the resource that shows how much computers can be used to help people to do work on behalf of people. We’re heading to that world, and it’s one that we’re all building collectively.

That would explain the massive investments that you’ve led making these big infrastructure bets…

Still not enough. We’re gonna feel the scarcity. We’re gonna feel it. We’re feeling it already. You can sense it now on people who are trying to use these agents, and just simply cannot, hitting the rate limits. So we’re working on behalf of our customers, on behalf of everyone who wants to use these agents to ensure that there is enough. And I don’t think we’re gonna get there. We’re gonna do our best, but I think that we are headed to a world of compute scarcity. And again, I think this is something where we can all contribute to trying to help there just be more availability of this in the world.

Greg, thanks again for coming on.

Likewise, great chatting.

Google’s rapid AI improvement has lifted its once-beleagured cloud division, with startups and enterprises now clamoring to build on top of the newly competitive Gemini. And this week, the company is set for its next big moment, as it brings together its stakeholders at the fast-growing Google Cloud Next event in Las Vegas.

Though Google Cloud Platform still trails its rivals in terms of market share, the business is growing fast, with Q4 cloud revenue increasing 48% to $17.7 billion. Meanwhile, Google’s cloud revenue backlog more than doubled in 2025, growing to $240 billion by the end of last year. Google’s also said AI customers on average use 1.8x as many Google products as non-AI customers.

Cloud Next, which has long outgrown San Francisco’s Moscone Center, will be a moment for the company to mark its progress in the cloud wars, a competition where it was once left for dead.

Given the AI race’s speed, it won’t be safe for Google to take a breath though. With the rise of Claude Code and OpenAI’s Codex, the company’s now under pressure to bolster its own agentic coding play. Google’s reportedly created a “strike team” to improve its AI coding models. And in the meantime, it will focus on helping its customers build agents of their own, a focus of its Vegas event.

Unlike some of the more messianic-style messaging you’d hear from Google’s competitors, the company is taking a more grounded approach this year. Bottlenecks, for instance, will be a key theme at Cloud Next, with more than a dozen sessions citing the topic. One session, “The human bottleneck: Why great tech fails and how to drive value in AI,” features World Labs co-founder Fei-Fei Li.

The bottleneck theme is pertinent given the ‘capability overhang’ that so many AI labs talk about. By this, they essentially mean that AI’s abilities have outpaced humans’ capability to implement them. If Google is betting that further growth will come from helping customers solve these problems, the payoff could be further gains against Amazon Web Services and Microsoft’s Azure.

Here’s a sample of this year’s Cloud Next talks:

• NVIDIA Omniverse For Physical AI: From Simulation to Reality

• The journey to agentic applications at Spotify

• Securing global innovation: How industry leaders scale defense across complex environments

• From algorithms to agents: How Gen Z, Gen Alpha, and AI are rewiring commerce

• Inside the race to build AI factories (Alex is participating)

The Intelligence Report

• The National Security Agency is reportedly using Anthropic’s new Mythos model despite the government blacklist.

• Google added new AI features for Chrome, including new search tools and “skills” powered by Gemini.

• Meta is expected to lay off around 10% of staff next month. Meanwhile, it’s also reportedly building an AI clone of Mark Zuckerberg to attend meetings and talk with staff.

• OpenAI had several major departures, including Kevin Weil (former chief product officer), Srinivas Narayanan (CTO of B2B Applications) and Bill Peebles (head of Sora).

• Publisher efforts to block AI scraping are creating serious worries for the Internet Archive, which operates the Wayback Machine.

• Amazon debuted a new agentic AI tool for drug discovery via AWS.

• Microsoft is reportedly testing a new always-on agent for Copilot similar to OpenClaw.

• A humanoid robot won a half marathon race in Beijing, beating the human race record by several minutes. (It was one of 300 robots that competed in Sunday’s race.)

• Emil Michael, Under Secretary of War for Research and Engineering at the Pentagon, joined the Big Technology podcast to talk about how AI is shaping modern warfare, his decision to label Anthropic as a supply chain risk, and what he thinks about the Pentagon Pizza Index.

Super Apps Vision Come Into Focus

Last week, OpenAI debuted a new Codex super app for desktops, with coding tools and other features for computer control and an in-app browser. The updates for Codex also include more than 90 extra plugins that combine skills, app integrations and MCP servers. OpenAI also released a new reasoning model called GPT-Rosalind to support life sciences research including biology, drug discovery and translational medicine.

Other AI model updates announced:

• Anthropic released a new AI tool called Claude Design that lets users create prototypes from text prompts. It’s powered by Claude Opus 4.7, which was also released last week.

• Perplexity released a new “Personal Computer” AI assistant for Mac users, with multi-model orchestration through new computer agents for working with files, apps, websites and other connectors.

• Google debuted a new Gemini desktop app for macOS, featuring a way to analyze content in any shared window. Google also released a new Gemini desktop app for Windows desktops.

7 Stats From Stanford’s 2026 AI Index

Stanford’s 2026 AI Index provides yet another annual deep-dive into AI’s technical progress, economic influence and societal impact. With more than 400 pages, the annual report from the Stanford Institute for Human-Centered AI features far more than we can fit into this section. But here are a few quick stats and other insights that stood out.

• AI is getting better at performing in professional domains like tax evaluations, mortgage processing, corporate finance and legal reasoning. (Accuracy for the top 15 models ranged between 60% and 90%.)

• The performance gap between U.S. and Chinese AI models is closer than ever, with the top U.S. models now only ahead by 2.7%

• Video models like DeepMind Veo 3 are getting better at generating content with realistic object behaviors without being trained on them. Two examples: simulating buoyancy and solving mazes.

• The number of AI researchers moving to the U.S. has slowed drastically, falling by 80% in just the last year.

• The number of AI safety incidents jumped to 362 in 2025, up from 233 in 2024.

• Open-source AI projects continue scaling, with 5.6 million projects on GitHub and Hugging Face uploading tripling in three years.

Jensen Huang, master of media, didn’t have it this time. The Nvidia CEO stumbled in an interview with podcaster Dwarkesh Patel this week, losing his calm veneer — while issuing an instant-classic “I didn’t wake up a loser” quote — and struggling to make satisfying arguments on his company’s two most critical issues.

Huang’s built Nvidia into the world’s largest company by making its technology essential to AI computing, but when pressed about competition and restrictions on selling into China, his answers fell flat. Nvidia stock didn’t flinch following the conversation (demand for its technology remains insatiable) but Huang’s responses didn’t exactly slam the door shut on its long-term challenges either.

Nvidia’s biggest threat today is well-funded competitors creating purpose-built AI chips that do a good enough job training and running AI models. Google’s Gemini and Anthropic’s family of models, including the latter’s Mythos, have been built outside of Nvidia’s ecosystem, a counterweight to Nvidia’s dominance. In the interview, Huang spelled out that Mythos, a model reportedly so powerful Anthropic won’t release it to the public, was built on basic computing hardware.

“Mythos was trained on fairly mundane capacity,” Huang said. “And a fairly mundane amount of it.”

Huang’s Mythos reference was meant to illustrate that China could build strong AI models without Nvidia technology (more on that shortly) yet highlighted that Nvidia isn’t a necessary component in frontier AI models today. If Anthropic could build Mythos on “mundane capacity,” then perhaps Nvidia isn’t a must-have for its counterparts.

Anthropic, to be sure, has struggled to meet demand using its current compute architecture. Its users often complain about Claude’s rate limits. And, per Huang, it’s almost solely responsible for Nvidia’s competitors’ growth. “Anthropic is a unique instance, not a trend,” Huang said. “Without Anthropic, why would there be any TPU growth at all? It’s 100% Anthropic. Without Anthropic, why would there be Trainium growth at all? It’s 100% Anthropic.”

Still, of the top three AI labs in the world, as Patel noted, only OpenAI is reliant on Nvidia’s GPUs. Meta and Grok, which have made massive GPU investments, aren’t faring as well as hoped. This doesn’t mean Nvidia is cooked, but it could cut into growth over time. And Huang was uncharacteristically flat-footed on that line of questioning.

The China Question

Export restrictions on sales to China is another meaningful threat that looms over Nvidia’s business, but Huang didn’t persuasively argue against them either. He simply refused to entertain the possibility that selling more powerful chips to China could enhance that country’s offensive cyberwarfare capabilities.

Asked repeatedly whether he would acknowledge the potential threat that could result from giving China more powerful chips, Huang evaded time and again. “I just want you to acknowledge that any marginal sales for the American technology industry is beneficial,” Huang retorted in one exchange.

Huang isn’t in the easiest position in these interviews. He must balance his relationship with the U.S. government, China’s leadership, his employees, the public, and shareholders. This may have hamstrung him from making the strongest case against export controls. That is: Exporting Nvidia’s technology worldwide means there’s a better chance American values will be baked into LLMs and their applications moving forward. Huang probably couldn’t say that outright, out of concern for China’s reaction. But he did hint at it in the strongest part of the interview.

“We get the benefit of American technology leadership,” Huang said of dropping export controls. “We get the benefit of developers working on the American tech stack. We get the benefit, as those AI models diffuse out into the rest of the world, that the American tech stack is therefore the best for it. We can continue to advance and diffuse American technology. That, I believe, is a positive.”

Huang’s interview has been something of a Rorschach test in the tech industry since its release. For some, he valiantly made the case for Nvidia against an unnecessarily adversarial interviewer. For others, he alarmingly lost his composure in the face of standard, informed questioning. The truth is probably somewhere in the middle: Nvidia remains a dominant, key driver (and beneficiary) of the AI boom. It’s also not invincible. No company is.

Subscribe now

An Autonomous AI Assistant for Everyone (sponsor)

You don’t need another productivity “hack.” You need real help you can deploy today.

Use Wingman, the fully-customizable AI agent that works where you do.

Wingman handles low-stakes tasks (social media, emails, research, ops) without monitoring, so you can focus on other priorities.

• Works inside WhatsApp, Telegram, iMessage, Gmail, etc

• Build one in 5 minutes (no coding or experience needed)

• Requires confirmation for anything consequential or sensitive

It’s private, secure, and designed to act with your permission.

Build your Wingman today.

Learn More

Advertise with Big Technology?

Reach 150,000+ plugged-in tech readers with your company’s latest campaign, product, or thought leadership. To learn more, write alex@bigtechnology.com or reply to this email.

What Else I’m Reading, Etc.

Anthropic asks religious leaders if Claude can be a “child of God” [Washington Post]

Dario Amodei is heading into the White House for peace talks [Axios]

Allbirds, somehow, is a GPU company now? [Bloomberg]

Google’s AI mode comes to Chrome [Techcrunch]

SantaCon leader accused of conning the santas [New York Times]

This Week On Big Technology Podcast: The Pentagon’s AI Plan + Behind the Anthropic Fight — With Under Secretary of War Emil Michael

Emil Michael is the Under Secretary of War for Research and Engineering at the Pentagon. Michael joins Big Technology to discuss how AI is transforming the Department of War, from targeting systems to drone warfare to cyber defense. Tune in to hear his account of why the Pentagon designated Anthropic a supply chain risk, what actually happened in the contract negotiations, and whether the decision was wise. We also cover how the military’s Maven Smart System works in practice, what the U.S. learned from drone warfare in Ukraine and Iran, and whether the Pentagon Pizza Index is credible. Hit play for one of the most candid conversations you’ll hear about AI and national security.

You can listen on Apple Podcasts, Spotify, or your podcast app of choice

Thanks again for reading. Please share Big Technology if you like it!

And hit that Like Button, you didn’t wake up a loser after all

My book Always Day One digs into the tech giants’ inner workings, focusing on automation and culture. I’d be thrilled if you’d give it a read. You can find it here.

Join Big Technology’s Private Discord Server!

Where we’ll talk about this story, the latest in AI, the week’s podcast, and plenty more. You can sign up via the link below:

Read more

After Anthropic’s introduction of Mythos — the mega-model it said is too dangerous to release to the public — OpenAI is poised for AI’s next big moment with “Spud.”

“Spud” is a massive new base model that OpenAI President Greg Brockman recently told us is “a new pre-train” that can understand instructions and context better. Brockman said Spud will be better “qualitatively” and “quantitatively” and better at solving “much harder problems.” Along with Mythos, it’s in a class of new, larger AI models whose performance will reveal just how much improvement is left in making AI models bigger.

“There’s this thing called ‘big model smell’ that people talk about, where when these models are just actually much smarter and much more capable, they bend to you much more,” Brockman said. “And you feel it — when you ask a question and the AI doesn’t quite get it, it’s always so disappointing. You have to explain it, you’re like: you really should be able to figure this out.”

Mythos, which Anthropic revealed last week, is a new general-purpose model the company says is too powerful to release publicly due to cybersecurity risks. It launched Project Glasswing — a closed program giving access to roughly 50 partners including Apple, Microsoft, Google, Nvidia, CrowdStrike, JPMorgan Chase, and others. Anthropic illustrated the risks by including an anecdote in Mythos’s system card about a time Mythos successfully escaped its sandbox, which a researcher only discovered after receiving a surprise email from the model while eating a sandwich in a park.

Framing AI models as too dangerous to release and available only to a select group is starting to look like a new trend. When OpenAI recently hinted at a separate cybersecurity product it’s developing, company officials said it also plans to release it only to a small partner group.

That raises a number of questions worth considering as these models come out: Has “most dangerous” simply become a new way of signaling “most powerful”? Is all the scary stuff mostly marketing? Is this danger talk a cover story for a lack of compute needed to deliver these models to the public? And who benefits most when access to the most capable models is concentrated among a handful of companies that already dominate the industry?

It’s also hard not to notice the names Anthropic and OpenAI are using for their next-generation models. Spud calls to mind something plain, unassuming, and earthy. Mythos resembles something much more lofty and lasting. Like any myth, there’s always a chance the story is grander than reality.

For now though, companies and governments are taking the companies’ claims seriously. After Anthropic announced Mythos, the Treasury and the Fed both reportedly warned major banks about new AI risks posed by Mythos’s capabilities. The head of the International Monetary Fund is also reportedly worried and said that “time is not our friend on this one.”

Subscribe now

Info-Tech LIVE 2026: Where CIOs Turn AI Momentum Into Measurable ROI (sponsor)

If you’re a CIO or senior IT leader, the pressure to show real business value from AI has never been higher. Info-Tech LIVE 2026 in Las Vegas puts you in the room with high-caliber IT executives solving enterprise-scale challenges and gives you the practical strategies you need to justify your AI spend and lead your organization through what comes next.

Submit your details through my link to claim your free ticket, subject to eligibility.

Learn More

Sponsored by Info-Tech Research Group

The Intelligence Report

• A Bank of America analyst raised their chip forecast to $1.3 trillion, which is $300 billion higher than four months ago.

• Meanwhile, a recent report from McKinsey suggests the semiconductor market could double to $1.6 trillion by 2030, up from $775 billion in 2024.

• AI inference is likely to drive more data center demand than training, according to McKinsey, which expects the former to surpass the latter by 2030.

• Meta announced its new Muse, the first in a new “Spark” family of models, marking the first release in a while from the company’s superintelligence team. The company also announced a new $21 billion deal with CoreWeave to expand cloud computing capacity.

• Anthropic said its run-rate revenue has surpassed $30 billion, more than 3x what it was at the end of 2025. The update came as the company announced a new agreement with Google and Broadcom for multiple gigawatts of compute

• Amazon CEO Andy Jassy used his 2025 shareholder letter to make a bullish case for the company’s AI ambitions, defending heavy AI spending, plans to “flatten” the organization, and outlining a vision for how companies should navigate the AI era.

• OpenAI published a new policy blueprint pitching a broad social contract around AI that includes workforce support, public wealth funds, and institutional reforms.

• The European Union is exploring whether ChatGPT should be considered a large online search engine under the Digital Services Act, which would make it subject to much stricter EU regulation related to AI and data privacy.

Federal and State Governments Zero In on AI

Data centers are facing growing pressure at the national, state, and local level. Congress, state lawmakers, and local governments are exploring ways to ban, pause and, otherwise regulate AI infrastructure amid concerns about rising electricity costs, environmental impacts and economic uncertainty.

• Last week, Maine lawmakers passed legislation that would create a moratorium on data centers larger than 20 megawatts through November 2027 and create a new state council to evaluate issues related to current and future potential data centers in the state. The bill still needs to be signed by Governor Janet Mills, who is also running for a seat in the U.S. Senate, where she’s been trailing by double digits.

• While Maine might be the first state, nearly a dozen other states are considering their own pauses or bans on data center construction, including Virginia, Georgia, New York and Vermont.

• Governors are also applying pressure. A bipartisan group of governors from Illinois, Delaware, Maryland, and Pennsylvania sent a letter last week to PJM Interconnection — the grid operator serving much of the mid-Atlantic and Midwest — calling for data centers to pay more for power and for stronger consumer protections. One of the signatories is Illinois Gov. JB Pritzker, who has separately proposed a pause on his state’s data center tax credit program.

• In Congress, Sen. Bernie Sanders and Rep. Alexandria Ocasio-Cortez introduced a new bill last month that would create a national moratorium on data centers with no fixed end date, keeping the ban in place until Congress passes a broader package of AI safety and labor protections. The bill would also ban the export of semiconductors, chips, and other computing hardware to countries that don’t have equivalent AI regulations in place — a provision that would have sweeping implications for U.S. trade policy.

This week, Congress returns from recess with committees in both chambers scheduled to discuss AI issues and other tech-related topics.

• On Wednesday, the House Education and Workforce Committee will meet to discuss AI’s economic impact on workers and employers. The hearing will be the latest in an ongoing series to understand AI’s impact on issues like work broadly, workplace safety, teaching in the AI age, and employer-led workforce training.

• The Senate Commerce, Science and Transportation Committee will meet on Tuesday for a markup of several tech-related bills.

• Two bi-partisan social media bills are on the agenda. One is the “Stop the Scroll Act,” which would direct the FTC, with Surgeon General approval, to require mental health warning labels on social media platforms. The other, the “No Fentanyl on Social Media Act,” would require the FTC to report to Congress on minors getting access to fentanyl on social media.

• The National Quantum Initiative Reauthorization Act of 2026 would accelerate quantum technology development and real-world applications. The bipartisan bill has already received endorsements from Google, Microsoft, IBM, the American Physical Society, and various quantum companies and academic institutions.

Big Technology Podcast Friday Edition: Anthropic’s Mythos Dilemma, Violence Against AI, Tokenmaxxing at Meta

Ranjan Roy from Margins is back for our weekly discussion of the latest tech news. We cover: 1) Anthropic’s new Mythos preview 2) Is Mythos marketing or a legit breakthrough? 3) The Mythos sandwich guy story 4) OpenAI and Anthropic’s brewing 1st party vs. API conflict of interest 5) The Meta-Harness 6) Violence against AI on the rise 7) Maine is going to pass a data center moratorium 8) Was Medvi really a $1.8 billion two person startup? 9) Tokenmaxxing is all the rage

You can listen on Apple Podcasts, Spotify, or your podcast app of choice

Join Big Technology’s Private Discord Server!

Where we’ll talk about this story, the latest in AI, the week’s podcast, and plenty more. You can sign up via the link below:

Read more

The wooden front door was filled with bullet holes.

There were twelve distinct entry points, all shot by a crazed opponent of data centers into the home of Indianapolis legislator Ron Gibson. The bullets landed just steps away from where Gibson’s 8-year-old son had built LEGOs a day before. On the front porch, the shooter left a handwritten note. “No Data Centers,” it said.

The shooting happened on Monday, and was mostly ignored in a week dominated by news of a ceasefire with Iran. But it was yet another data point — and definitely the most disturbing one — indicating a growing, serious unease with AI and its associated physical infrastructure.

Today’s AI builders are in a bind: They need to massively increase the quantity of data centers to improve their models, but as their models get better, the public’s support for their buildouts is eroding.

As data centers show up in local communities, far more Americans say their impact is more negative than positive for quality of life, the environment, and home energy costs, according to Pew Research. Only 6% of those polled said AI infrastructure has a positive effect on the lives of people nearby. The faceless, imposing buildings have become symbols of the tech industry’s interest in delivering AI quickly despite the uncertainty it will bring to people’s lives.

Share

Already, some legislators are introducing bills to stop data center construction, and some are close to going through. Maine, for instance, is en route to become the first state to pass a data center construction moratorium through November 2027. And though unlikely to pass, Senator Bernie Sanders and Congresswoman Alexandria Ocasio-Cortez are testing the appetite for further restrictions on a national level with a bill to pause data center expansion in the U.S.

Ilya Sutskever, the OpenAI co-founder and current CEO of Safe Superintelligence, once said that when the race toward AGI breaks out, data centers could wallpaper the planet. “It’s pretty likely,” he said, “the entire surface of the earth will be covered with solar panels and data centers.”

But the public has at least some say in how its resources will be used. And, though investment continues to pour in, achieving Sutskever’s vision will not be as simple as throwing dollars at the problem.

Any political opposition to the AI buildout would further complicate these ambitious infrastructure projects, which are already struggling to stay on schedule. Today, there’s concern that as many as half the scheduled data centers meant to come online this year could be delayed. Power and equipment constraints are playing a key role in the delays, but rising political opposition could loom large as the projects attempt to scale up.

Delivering AI is a famously resource-intensive process. And now, just as the AI models are showing their most promise yet, the public is turning against the foundational part of the stack (with a chance to impede their progress in a way they never could with software or social media). Ultimately, the technology will likely win out. But the road to further progress could be much bumpier than anticipated.

Subscribe now

Apple’s Starlink Update Sparks Huge Earning Opportunity (sponsor)

Apple just secretly added Starlink satellite support to iPhones through iOS 18.3.

One of the biggest potential winners? Mode Mobile.

Mode’s EarnPhone already reaches 490M+ users that have earned over $1B, and that’s before global satellite coverage. With SpaceX eliminating “dead zones,” Mode’s earning technology can now reach billions more in unbanked and rural populations worldwide.

Their global expansion is perfectly timed, and investors like you still have a chance to invest in their pre-IPO offering at $0.50/share.

With their recent 32,481% revenue growth and newly reserved Nasdaq ticker, Mode is one step closer to a potential IPO.

Tap into a $1T opportunity — invest now at just $0.50/share and get up to 20% bonus.

Learn More

Disclaimer

Please read the offering circular and related risks at invest.modemobile.com. This is a paid advertisement for Mode Mobile’s Regulation A+ Offering.

Mode Mobile recently received their ticker reservation with Nasdaq ($MODE), indicating an intent to IPO in the next 24 months. An intent to IPO is no guarantee that an actual IPO will occur.

The Deloitte rankings are based on submitted applications and public company database research, with winners selected based on their fiscal-year revenue growth percentage over a three-year period.

This Week On Big Technology Podcast: OpenAI vs. Anthropic’s Direct Faceoff + Future of Agents — With Aaron Levie

Aaron Levie is the CEO of Box . Levie joins Big Technology to discuss the battle between OpenAI and Anthropic as their product roadmaps converge around coding, enterprise, and AI agents. Tune in to hear where AI agents are actually gaining traction, why coding has emerged as the breakthrough use case, and what stands in the way of broader adoption across knowledge work. We also cover trust and security concerns, the messy reality of enterprise data, and the debate over whether value will accrue to the labs or the application layer. Hit play for a sharp look at where the AI market is heading and who stands to benefit most.

You can listen on Apple Podcasts, Spotify, or your podcast app of choice

Thanks again for reading. Please share Big Technology if you like it!

My book Always Day One digs into the tech giants’ inner workings, focusing on automation and culture. I’d be thrilled if you’d give it a read. You can find it here.

Join Big Technology’s Private Discord Server!

Where we’ll talk about this story, the latest in AI, the week’s podcast, and plenty more. You can sign up via the link below:

Read more

OpenAI is in the midst of yet another strategic shift. The company has abandoned video generation, it’s building a ‘superapp’ that combines coding, chat, and browsing, and it’s zeroing in on a use case where AI uses your computer to make you more effective at work, and assist you everywhere else.

Underlying the shift, OpenAI President Greg Brockman told me, is a belief that the company’s core large language models are the right architecture to bet on, and that tree of AI is growing increasingly capable and reliable.

“There’s been this debate about how far text models can go,” Brockman said. “I think we have definitively answered that question — it is going to go to AGI. We have line of sight to much better models coming this year.”

In an extended Big Technology Podcast interview, Brockman offered revealing comments about the company’s research direction, how far it can push its Codex coding assistant, and the logic of supporting all of this with mountains of compute.

You can read the full Q&A below, edited lightly for length and clarity, or listen on Apple Podcasts, Spotify, or your podcast app of choice.

Alex Kantrowitz: OpenAI has shut down video generation and is preparing a ‘superapp’ that will combine business and coding use cases. From the outside, it looks like you were winning in consumer but decided to shift toward business. Why do that?

Greg Brockman: We’ve been in a world where we’re developing this technology, deep learning, to really see, can it have the positive impact that we have always pictured? And we’ve separately had an effort to actually try to deploy this technology, whether that’s to help sustain the business, to start getting some practice with getting real world impact, those kinds of things.

We’re at a moment now where we’ve really seen this technology is going to work, and that we’re moving out of testing on benchmarks and these almost cerebral demonstrations of capability to it actually being the case that for us to develop it further, we need to see it in the real world and get feedback from how people are using it in knowledge work, in various applications. And so the way I think about it is that this is a bigger strategic shift because of the phase of the technology.

It’s not so much that we’re saying we’re moving from consumer to B2B, what we’re saying is — What are the most important applications that we can focus on, because we can’t focus on everything. What are the things that we can bring to life that will actually synergize together as we build them, and that will deliver meaningful impact and help elevate everyone?

When we look at the list, there’s consumer, which you can think of it as many things, but there’s a personal assistant — something that knows you, that’s aligned with your goals, it’s going to help you achieve whatever it is that you want in your life. There’s also creative expression and entertainment and many other applications. On the business side, maybe if you zoom out, it looks more like: You have a hard task, can AI go do it? Does it have all the context to do all these things?

For us, it’s very clear that the stack rank includes two things at the top. One is the personal assistant, the other is the AI that can go and solve hard problems for you. And when we look at the compute we have, we are not even going to have enough compute to fund those two things. And then once we start adding in many other applications, many other things that AI is going to be very useful for and is going to help people with, we just can’t possibly get to all of them.

So this is a recognition of the maturation of the technology and the incredible impact it’s going to have very quickly, and our need to prioritize and to actually pick the set of applications that we want to really bring to the world.

I’ve heard you compare OpenAI’s various bets to Disney’s, where you have a company with a core advantage that it farms out in different ways. Disney has Mickey Mouse, and it can do movies, theme parks, Disney+. At OpenAI, you have the model, and you can do video generation, a personal assistant, enterprise work. Is that no longer possible?

In some ways, that story is even more true than it’s been. But the thing that’s important to realize is technologically, the Sora models — which are incredible models, by the way — are a different branch of the tech tree than the core reasoning GPT series. They’re just built in a very different way. And to some extent, we’re really saying that pursuing both branches is very hard for us to do for these applications.

We are actually continuing the Sora research program in the context of robotics, which I think is very clearly going to be a transformative application, which is still a little bit in the research phase.

And so it’s a recognition that for this moment, we really need to put the primary focus on developing the GPT series. And that doesn’t just mean text. It doesn’t just mean cerebral things. For example, bidirectional communication, having a great speech to speech interface — that is something that also is going to make this technology very usable and very useful. But it’s not a different branch of the tech tree. It’s all one model, and we just sort of tweak that in slightly different ways.

If you branch too far and you have two different artifacts, it is very hard to sustain in a world where there is limited compute, and the reason there’s limited compute is because there’s so much demand. There’s so much that people want to do with every single model that we create.

Betting on Text vs. World Models

Why is your bet on the GPT model tree, when you had been seeing real progress with Sora?

The problem in this field is too much opportunity. The thing that we observe very early on in OpenAI is that everything we could imagine works now. There’s different levels of friction associated with it, different amounts of engineering effort, different compute requirements, all those things, but every single different idea, as long as it’s kind of mathematically sound, you actually can start getting some pretty good results.

That shows you the power of the underlying technology of deep learning, the ability to really take any sort of problem and to get to the meat of it, to have an AI that really understands the underlying rules that generated the data. So it’s not about data itself, it’s about understanding the underlying process and then being able to apply it in new contexts. So you can do that in world models. You can do that in scientific discovery. You can do that in coding.

And I think that where we are as we think about the rollout of this technology is that there’s been this debate of, how far will the text models go, how far can text intelligence go? Can you have a real conception of how the world operates? And I think that we have definitively answered that question — it is going to go to AGI. We see line of sight. And at this point we have line of sight to much better models that are coming this year, and the amount of pain within OpenAI that we’ve had to decide how to allocate compute goes up, not down, over time.

And so I think that maybe the core of it is that it’s about sequencing and timing, and that in this moment, the kinds of applications that we’ve always dreamed of are starting to come into reach. For example, solving unsolved physics problems — we had this result recently where a physicist had been working on a problem for some time, he gave it to our model, 12 hours later we have a solution. And he said this is the first time he’s seen a model where he felt like it was thinking, that it felt like this is a problem that maybe humanity would never solve, and our AI solved it. When you see something like that, you have to double down, you have to triple down, because we can really unlock all of this potential for humanity.

And so for me, it’s not about relative importance of these things. It’s more about what is OpenAI’s mission of delivering AGI to the world, our vision of how it can benefit everyone, and the fact that we have a tech tree that we see how to just push it, how to do the engineering, do the further science and research to then have that come to fruition.

Does open AI potentially miss something by doubling down on the text model tree?

So two answers. One is: absolutely, yes. In this field, you do have to make choices. You have to make a bet, and that’s actually where OpenAI started — we really said, what is the path to AGI that we believe in? And really focused hard on that. The sum of random vectors is zero, but if you align your vectors, then you can go in a direction.

But the second point is — it’s actually image generation that has been very, very popular within ChatGPT, and that’s something we’re continuing to invest in, continuing to prioritize. And the reason we’re able to do that is because it’s not actually on the world model, diffusion model tech branch — it’s actually based on the GPT architecture. And so there, even though it’s a different data distribution, the actual core technology, the core stack, it’s all one thing.

That is actually the pretty wild thing about what AGI is, is that sometimes these very different looking applications — between speech to speech, image generation, text — and text is, by the way, itself many facets, like science and coding and personal wellness information, those kinds of things — all of that you can do in one technological envelope. And so a lot of what I’m looking at, and what we as a company are looking at from a technological perspective, is how to have as much unification of our efforts, because we really see this technology as being something that’s going to uplift and power the whole economy. The whole economy is a massive thing, and so we can’t possibly do all of it, but we can do our part.

That’s the general part in artificial general intelligence. That’s the G.

That’s the G…

Our conversation continues below…

Should Software Companies Embrace AI or fight it? — With Asana Chief Product Officer Arnab Bose (sponsor)

Arnab Bose is the chief product officer of Asana. Bose joins us to discuss whether software companies should embrace AI or fight it as agents begin to reshape how work gets done. Tune in to hear how Asana is thinking about AI Teammates, the future of work management, and what onboarding AI into an organization looks like in practice. We also cover Asana’s work with leading model providers, the case for and against open source, and what widespread agent adoption could mean for labor. Hit play for a smart conversation about whether AI strengthens software companies like Asana or threatens to make them obsolete.

Watch Now

OpenAI’s Super App

Speaking of unifying things. What is this ‘super app’ going to be? It’s going to bring together coding, browser, and ChatGPT?

That’s right. We want to build an endpoint application for you that really lets you experience the power of AGI — the generality. Think about what chat is today. I think chat is really going to become your personal assistant, your personal AGI. An AI that’s looking out for you, that knows a lot about you, that’s aligned with your goals, that’s trustworthy, that kind of represents you in this digital world.

Codex you can think of as — right now it’s been a tool that we built for software engineers, but it’s becoming Codex for everyone, that anyone who wants to build can use Codex and to get the computer to go do the thing that they want. And it’s not just about the actual software anymore. It’s really about almost the use of a computer — whether it’s to set up, I used to set settings on my laptop, I forget how to set the hot corners. You ask Codex to do it, it just does it. That’s what computers were always supposed to be — contort to the human, rather than the human contort to them.

And so imagine one application where anything you want your computer to do, you can ask it. There’s computer-use browsing built in, for an AI to be able to actually use a web browser, and for you to be able to oversee what the AI is doing. All of your conversations, regardless of application — whether it’s for chat or whether it’s for code, whether it’s for general knowledge work — that’s all unified in one way, so that the AI has memory, knows about you. That is what we are building.

But it’s really an iceberg, because that’s the tip. What to me is actually much more important is the technological unification. The thing that’s really changed over the past couple of years has been that it’s no longer just about the model. It’s about the harness. It’s about how does the model get context? How is it connected to the world? What actions can it take? How does the loop of interacting with the model work as you get new context?

All of that was something that we had multiple implementations of, slightly different, and we’re converging it. We’re going to have one version of that, and almost end up with this AI layer that can be pointed at specific applications in a very thin way. So you can build a little plugin, a little UI if you really want something that’s great for finance, great for legal, but you generally won’t have to — because this one super app will be very broad.

This app is for business use cases, personal use cases?

Your laptop – is it for personal? Is it for business? It’s both. It’s your personal machine that gives you an interface to this digital world, and that’s what we want to build.

So just talk a little bit about from a non-business standpoint — I’m using the super app in my personal life. What am I using it for? How does my life change?

I would think of it as: personal life, just the way that you use ChatGPT right now. And people use it for such a diversity of really amazing applications — sometimes that’s just asking, I’m going to give a speech at a wedding, can you help me with drafting it? Can you give me some feedback on this idea that I have? I’m working on a small business, can you give me some ideas there? Which maybe starts to bridge between personal and work. Any of those questions should be things that you can go to the super app for and it answers.

But if you think about what ChatGPT has been, it’s already been evolving. It used to not have any memory — it’s just the same AI for everyone starting from scratch. It’s almost like talking to a stranger. It’s way more powerful if it remembers the interactions you’ve had. It’s way more powerful if it has access to context — if it’s hooked up to your email and to your calendar, and really knows your preferences and has this almost deeper set of past experiences with you that it’s able to leverage to achieve your goals. Pulse is a feature in ChatGPT right now where every day, it surfaces for you things that you might be interested in based on what it knows about you. So in the personal capacity, the super app will be doing all of that, and will be doing it in a much deeper and richer way.

When are you planning to ship it?

We’re taking incremental steps to get there. Over the next couple of months we should have shipped the complete vision of what we’re talking about here, but it’s going to come in pieces.

The place that we’re starting is with, for example, the Codex app today — which is really two things in one. It’s a general agent harness that can use tools, and it’s also an agent that knows how to write software. That general agent harness can be used for so many different things. You hook it up to spreadsheets, you hook it up to Word documents, it’s able to help you with knowledge work. And we’re going to make the Codex app just so much more usable for general knowledge work, because what we’ve already seen within OpenAI is all this organic adoption of people using it for that. So that’ll be the first step, and there are many to come.

Codex’s Potential Beyond Software Engineering

I was speaking with one of your colleagues yesterday taking a look at Codex, and he mentioned that someone using Codex had instructed it to help them with video editing. It builds a plugin for Adobe Premiere, started separating it into chapters and started the edit.

I love hearing that. That’s exactly the kinds of things that we want this system to be useful for.

The Codex app itself was originally built for software engineers, and the current usability of it for non-software engineers is actually quite low, because there’s a bunch of little things where when you set things up, you run into some error that a developer knows what it means, knows how to fix it — it’s just kind of what we’re used to. But if you’re not a developer, you’re like, what is this? This is not something that I’ve encountered before. And despite that, we are seeing people start to use this who have never programmed before, to be able to build websites, to be able to do exactly the kinds of things you said — to be able to automate their interactions with different pieces of software, to be able to get lots and lots of leverage. Someone on our communications team uses it hooked up to Slack and to their email — they’re able to go through a bunch of feedback and synthesize it very well.

People who are very motivated can jump through the hoops and then get great return from it. We did the super hard part of building an AI that is really smart, capable, and can actually accomplish your task. Now we have to do the much easier part, in some sense, of making it broadly useful and removing these barriers to entry.

Looking at the competitive landscape, Anthropic has the Claude app. You can use Claude the chatbot, Claude Code, so they have a version of a super app of their own. What do you think Anthropic saw that got them to this position earlier, and what do you think your chances are of catching up there?

If you rewind 12, 18 months, we have always been focused on coding as a domain. We always had the best numbers on different programming competitions, these very cerebral things. But the thing that we didn’t invest in as much was that last mile of usability, of really trying to think about — okay, this AI is so smart it can solve all these great programming competitions, but it’s never seen someone’s real world codebase, which is messy and not quite as pristine as the world that it sort of has experienced. And I think that is something that we were behind on.

But about maybe mid last year is when we got very serious about that, and we had a team very focused on: what are all the gaps, what are all the kinds of messiness of the real world that we haven’t encountered? How do we actually get training data, build training environments that let the AI experience what it’s like to actually do software engineering, be interrupted in weird ways, all those things?

I’d say at this point we are caught up — when people go head to head, us versus competitors, people tend to prefer us. We know we’re digging in on front end, we’re going to fix that.

But this is the general motion that we’ve been taking — to say that usability, thinking about the product end to end, not just a model and then building a separate thing, but really thinking of it as one product. When we’re doing the research, we’re thinking about how it will be used. That has been a motion that we’ve been changing within OpenAI. And so the way I would look at it is that we have incredible step-up models coming all year — I look at the roadmap, it’s truly inspiring, what will be possible. And then we’ve been really focusing now on: let’s also get the last mile usability.

You just used the phrase ‘we’re caught up.’ Is there a different vibe within the company, whereas now, instead of being the one that’s far ahead, you’re in a real fight?

For me personally, the scariest moment at OpenAI was actually after we launched ChatGPT. And I remember being at the holiday party and just feeling this vibe of: we won. I have never felt that. I was like: No. We are the underdog, and we always have been. The competitors in this space are established companies that have just so much more capital, so much more human resources, data, the whole thing. Why is OpenAI able to compete at all? And to some extent, the answer is only because we never feel complacent, where we always feel like we are the challenger.

And it actually, for me, has been a very healthy thing to see us start to see that in the marketplace, to see other competitors emerge and do a good job. In my mind, you can never fixate on your competitors. If you focus on where they are, then you’ll be where they are, and they’ll already have moved. And I think that’s what’s been happening in the other direction — a lot of people focus on exactly where we are and we get to move.

I’d say that the world that we’re in is one where you’re never as good as they say you are, you’re never as bad as they say you are. It’s been very steady. The core of the model production — that is something where I actually feel extremely confident in our roadmap and the research investments we’ve been making. And on the product side, we have such great energy that’s all coming together to deliver this to the world.

Forthcoming GPT Models: What’s Spud?

You’ve foreshadowed a couple of times already that you have some good models on the way. What is Spud? The Information said Sam Altman has told the staff the team believes this new “Spud” model can really accelerate the economy. What is it?

It’s a good model, but I think that it’s really not about any one model. The way that our development process works is you have pre-training, so you produce a new base model that is then the foundation that we build further improvements on top of. And that is always a huge effort across many people in the company — that’s where I’ve actually been spending most of my efforts over the past 18 months, really focused on our GPU infrastructure, on supporting the teams that do all of the training frameworks to scale up with these big runs.

But then there’s a reinforcement learning process. So you take this AI that has learned lots of things about the world, and it applies that knowledge, and then we do a post-training process where you really say: okay, now you know how to solve problems, you practice it in all these different contexts, and then here’s the last mile of behavior and usability.

So I think of Spud as a new base, as a new pre-train. I’d say it’s like we have maybe two years worth of research that is coming to fruition in this model. It’s going to be very exciting. And the way that the world will experience it is just improved capabilities. For me, it’s never about any one release, because as soon as we have this one release, it’ll be an early version of what we have coming. We’ll do much more of each of these steps of the improvement process. And so where we’re going is we have this engine of progress that just moves faster and faster, and Spud is just one step along the way.

What do you think it’ll be able to do that today’s models can’t?

I think it’s going to be able to solve both much harder problems. I think it will be much more nuanced. It’ll understand instructions better, understand the context much better. There’s this thing called ‘big model smell’ that people talk about, where when these models are just actually much smarter and much more capable, they bend to you much more. And you feel it — when you ask a question and the AI doesn’t quite get it, it’s always so disappointing. You have to explain it, you’re like: you really should be able to figure this out.

And so I would think of it as: quantitatively, lots of shifts, and qualitatively, there will just be new things where you would have been frustrated before, you’d never use AI for it, and now you just use it without thinking very much. I’m super excited to see how it raises the ceiling — we’ve already seen these physics applications and things like that, and I think we will be able to just solve way more open-ended problems over way longer time horizons. And I’m also very excited to see how it raises the floor, where just for anything you want to do, it’s just so much more useful for you.

It can be kind of tough for everyday users to really feel the change. There was a lot of buildup before GPT-5 came out, and actually the initial reaction was somewhat disappointment among the public. But then I think people realized that for certain tasks it was really good.

With these next series of models, do you expect that it’ll really be felt in the trenches in certain occupations? Or do you think it will be a broadly tangible improvement for everyone?

I think that it will be a similar story — when you release it, there will be people who will try it and be like, this is night and day different from anything I’ve seen. And then there will be some applications where we weren’t necessarily intelligence-bottlenecked. And so if you have a model that’s more intelligent, maybe you won’t feel it right there.

But I think over time you will feel it, because the fundamental thing that shifts is how much do you rely on the system. We all interact with AI by having some mental model for what we think it can do. And that mental model shifts actually fairly slowly — as you get more experience, it does something magical for you, you’re like: oh wow, it can do that, I never imagined that.

We see this, for example, in applications like access to health information. I have a friend who used ChatGPT to understand different treatments for his cancer. He was told by doctors that he was terminal, that there was nothing they could do for him. He used ChatGPT to actually research a bunch of different ideas, and he was able to get treatment that way. And that’s something where you need to have some level of belief that the AI is going to be helpful in that application, for you to really put in the effort to get something out of it. And I think what we’re going to see is that for any application like that, it’s going to become so much more evident to everyone that the AI can help you. So it’s a little bit of the technology getting better, but it’s also our understanding of the technology shifting and catching up to that.

AI Takeoff

You have an automated AI researcher in the works, it’s supposed to come out this fall. What is that?

So the direction of travel right now: we are in this early phase of takeoff of this technology.

What does takeoff mean?

Takeoff is as the AI gets better and better on this exponential — and in part because we can use the AI to make the AI better, so our development process speeds up. But I also think when I think of takeoff, it’s also about real-world impact. In some ways, every technology is an S curve. Or if you zoom out, some of those S curves end up being an exponential. And I think that’s what we’re encountering right now. So the technology development is moving with increasing speed, and it’s this engine that’s picking up momentum. But it’s also in the world — there’s all of these tailwinds, because there’s chip developers that are getting more resourcing into their programs. There’s this economy of people who are building on top of it, trying to figure out how it fits into every different application. And all of that energy is just accumulating more and more into this takeoff phase of AI, becoming just a kind of sideshow to being the main driver of economic growth.

And so the researcher will be a moment where the AI — which we’re building right now — is doing a larger percentage of tasks that can run autonomously. And that doesn’t necessarily mean that we just let it off on its own and come back later and see if it did something good. We are going to be very involved in managing it — just like right now, if you have a junior researcher and you leave them on their own too long, they’re probably going to go down a path that’s not very useful. But if you have a senior researcher, or someone who has a vision, they don’t even necessarily need to know the mechanical skills — they will be able to provide feedback, review the outputs that the intern is producing, and provide direction in terms of the vision of what they want accomplished.

And so I think of this as a system that we’re going to build that will massively accelerate our ability to produce models, to make new research breakthroughs happen, to be able to make these models more useful and usable in the real world, and to do that at increasing speed.

So what’s it going to do? Are you going to say: go find AGI?

I think the way I think of it is something like that, to first order. And at a practical level, I would view it as taking the full end-to-end of what one of our research scientists does, and being able to do that in silicon.

Another way to think about takeoff is that progress in AI goes from incremental to gathering momentum and then sort of this unstoppable march to an intelligence that’s smarter than humans. Do you worry that there are possibilities for that process to go wrong?

Absolutely yes. I think that the way to get the benefits of this technology is also to really think about the risks. And if you look at how we’ve approached technology development from a technical perspective, we invest a lot in safety, security. A good example of this is prompt injections — if you’re going to have an AI that is very smart and very capable, hooked up to lots of tools, you want to make sure that it can’t be subverted by someone giving it a weird instruction. And that’s something that we’ve invested in quite a lot, and I think have really incredible results. We have an incredible team working on it.

It’s interesting to think about some of these problems where you can make analogies to humans — humans are also susceptible to phishing attacks, to being deceived in different ways, to not really understanding the full context of what they’re working on. And we bring those analogies into our development process and think about: whenever we develop a model, how do we ensure that it’s going to be aligned with people and be able to actually be helpful?

And that is something that we care quite a lot about. I think that there are bigger questions about the world, the economy, how does everything change, how does everyone benefit from this technology — they’re not purely technical, not purely something that OpenAI on our own will be able to solve. But yes, I think quite a lot about not just pushing forward the technology, but also really about: how do we ensure that we have the positive impact that is its potential?

Many of your counterparts have said, if everybody agrees to stop AI progress, we’ll stop it — and yey it doesn’t seem like it’s going to slow at all. Is the reward worth the risk?

I think the reward is worth the risk, but I think that is too coarse-grained of an answer in some sense. The way that I think about it is that we’ve asked from the beginning of OpenAI: what does a great future look like? How can this technology really be something that uplifts everyone? And you can think of there almost being two different angles. One is the centralization view of saying that, well, the way to make this technology safe is that you have only one actor building it, and so then you don’t have any pressures — you can really think about getting it right, and then figure out how to roll it out to everyone when it’s ready. That’s a pretty tough pill in some ways.

And I think that there’s a lot of properties that you can instead think about approaching differently, which we refer to as resilience. To think of it as this open system where there are lots of players who are developing the technology, but it’s not just about the technology — it’s about building societal infrastructure that helps this technology really go well.

If you think about how electricity has developed, that’s something where lots of people produce it, it actually has dangers and risks, but we all build our safety infrastructure in a diversity of different ways — around safety standards for electricity, around different ways of harnessing it, about how you scale it. There are regulations at these massive scales. Lots of people are able to use it in a democratized fashion. There are inspectors. There’s a whole system that’s been built around the needs of that technology, the proclivities of that specific technology.

And I think that one thing that we have really seen with AI is that it is something where we need this broad conversation. We need lots of people to be aware — if this technology is going to come and change everything for everyone, people need to participate in that. It can’t be something that’s done in secret by just one sort of centralized group. And so this has been, to me, a very core question to how this technology should play out, and something we really believe in — this resilience ecosystem that should emerge around the development of this technology.

70-80% to AGI

So you said we’re in takeoff. Nvidia’s CEO Jensen Huang said recently that he believes AGI has been achieved. Do you agree?

I think that AGI has a different definition to many people, and there are many people who would say that what we have right now is AGI. I think you can debate it. But maybe the thing that’s interesting is that the technology we have right now is very jagged — it is absolutely superhuman at many tasks. When it comes to writing code and those kinds of things, AI can just do it, and it really removes a lot of the friction to creating things. But there are some very basic tasks that a human can do that our AI still struggles with. And so it’s almost: where do you draw the cut line? It’s a little bit more of a vibe than a feeling, than it is science at the moment.

For myself, we’re definitely going through that moment. And if you were to show me five years ago the systems we have today, I’d have said: oh yeah, that’s what we’re talking about. But it’s just so different from anything we ever pictured. And so I think we need to adjust our mental models appropriately.

So you’re not there yet.

I’d say I’m basically like 70-80% there. So I think we’re quite close. I think it’s extremely clear that we are going to have AGI within the next couple of years in a way that is still going to be jagged, but where the floor of tasks — for almost any intellectual task of how you use your computer — the AI will be able to do that. Right now I have to give a little bit of an uncertain answer, because it’s almost like an uncertainty principle kind of thing — you can debate it. For my own personal definition, I think we’re almost there, and with maybe a little bit more, we will absolutely be there.

Greg, what happened in December 2025? It seems like it was an inflection point where all this idea of letting the machine code for hours uninterrupted went from theory to a moment where everyone said, I think I can trust this to keep going for a while.

New model releases really went from the AI being able to do like 20% of your tasks to like 80%, and that was this massive shift, because it went from being kind of a nice thing to do to — you absolutely need to retool your workflow around these AIs.

And for myself, I’ve very much had this moment where I have a test prompt that I’ve been using for years — build a website for me. I’d actually built this website back when I was learning to code, took me months. Over the course of 2025, it used to take like four hours and a bunch of different prompts to get it right. In December: one shot, just asked the AI one time, and it produced it, and did a great job.

So how did those models make the leap?

A lot of it is about the better base models. One thing about OpenAI is that we’ve been working on improving our pre-training technology for quite some time, and in that moment we got to see a little taste of what is going to be coming for the rest of this year. But it’s also really about not any one thing — we’re constantly pushing on every single axis of innovation.

The thing that’s very interesting about these models is that in some ways you get these leaps, and in some ways it’s all continuous. It didn’t go from 0% to 80% — it went from 20% to 80%. So in some ways it just got better. And I think we’ve actually seen this improvement continue with every single point release that we’ve had.

One of my engineers I work with very closely went from: he couldn’t get it to do the low-level, hardcore systems engineering he does, to now — he gives it a design doc, it actually implements it, adds metrics, observability, runs the profiler, improves it to the point that it’s the exact thing that he was hoping to produce. And so the way to think about it is: slowly, slowly, slowly — all at once. But it is all indicated by what’s working right now. Certainly within a year, sometimes much sooner, it’s going to be incredibly reliable.

And it’s surprised you, because I heard you talking not long ago about how Codex — this autonomous coder — was just for software developers. And earlier in this conversation, you said actually everyone can use this stuff. What led to your changed perspective on it?

I’d been focusing on Codex — it’s got the ‘code’ in it — as really being for coders, and thinking about people within OpenAI, because many of us are software engineers building for ourselves. It’s very natural to think that way.

But as this technology has been progressing, we’ve started to realize that the underlying technology we produced is mostly not about code at all. It’s mostly about solving problems. It’s mostly about being able to manage context and harnesses and think about how an AI should integrate and do work. And that’s something that becomes — even for code — suddenly anyone can have access, because you can manage something that’s going to go do work. If you have a vision, something you want to accomplish, you can describe your intent, the AI can execute, can get that done. But then it also starts to be like: why am I just focused on coding? There’s so much very mechanical skill associated with Excel spreadsheets, with presentations. And if the AI has the context, it has the raw intelligence now to be able to do these things at a great level. So if we can just make it more accessible, suddenly it goes from Codex is for coders to Codex is for everyone.

The OpenClaw Vision

You talked a little bit about the AI as something that will help run your life for you. Is that the logic for bringing the OpenClaw team in house?

The core thing about this technology is that figuring out how it’s useful, how people want to use it, what is the vision for agents, how it’s going to slot into people’s lives — that is a hard problem. And one thing I’ve seen across many generations of this technology is the people who really lean in, who have a lot of curiosity, who have a lot of vision — that’s a real skill, and that’s an emerging, very valuable skill in this new economy.

And Peter, who is the OpenClaw founder, is someone who’s got incredible vision and incredible creativity. And so to some extent it’s about the specific technology, but to some extent it’s really about: how do we take these capabilities and figure out how they slot into people’s lives? As a technologist, it’s very exciting, but as someone who is focused on bringing utility to people, that’s something that we are doubling down on and investing quite a lot in.

You had a pretty interesting quote about this recently, talking about getting these autonomous AI agents to work on your behalf. You said you become this CEO of a fleet of hundreds of thousands of agents that are completing your objectives, your goals, your vision, and you’re not in the weeds on exactly how different things are solved. And in some ways, this new way of work can make you feel like you’re losing your pulse on the problem. Is that good?

I think there’s a mixed bag. We need to acknowledge the strengths of what these tools can deliver and mitigate the weaknesses. Giving people leverage and agency, making it so that if you have a vision, something you want to accomplish, you can have a fleet of agents that will go do it for you. But if you think about how the world works, at the end of the day there’s an accountable party. If you’re trying to build a website, and your agent messes it up, and your user is affected — it’s not really the agent’s fault. It’s your fault. And so you need to care. And for people to use these tools right, you need to realize that human agency, human accountability — that’s a core part of the system. How the human uses the AI — that’s something that is deeply fundamental. And so the important thing is that as a user of these agents, you cannot abdicate responsibility. You cannot just say: the AI is just going to do stuff.

But you said feeling like you’re losing your pulse on the problem itself — that’s different than accountability.

To me they actually are linked together. Because the point is that if you’re a CEO, and you’re too far from the details — you’re running this company, this team, and you’ve lost your finger on the pulse — that is something that’s not going to lead to great results. And so the point I was trying to make is not that it’s a desirable thing for humans to not know about what’s going on.

There are some details that you probably don’t need to worry about, because you can trust they’ll be taken care of — like if you are working with a general contractor to build a house, there’s a bunch of details there that you probably don’t need to worry about. But at the end of the day, if there are details that are wrong, you should care about it. You should be aware. And so you cannot just blindly say I’m okay with losing my finger on the pulse. You need to lean in and say: I need to keep it there, to really understand the strengths and weaknesses. And as you disengage from some of these lower-level mechanical things, you should do it because you have built trust with a system that it will do a good job.

One last question about the models. You talked a little bit about the evolutions that the models have gone through — pre-training and fine-tuning, reinforcement learning that gets it more equipped to solve problems step by step and go out on the internet and do things. And now we’re in this moment where the models have learned through that process to use tools. What is next in that progression?

The world that we’re in is one of this increasing capability and depth of what the machine can do. Some of this is about: we’ve got tool use, but now we also need to actually build really great tools. Think about something like computer use — an AI that can actually use a desktop — then it is really able to do anything that you can do. But we also have to build a little bit for the machine to think about: how does enterprise credentialing work? How do audit trails and observability work? There’s a lot of technology to build to catch up with what the core model capability is.

And I think the overall direction of travel includes things like a really great speech interface, so you can just talk to your computer naturally — as natural as this conversation — and it understands you, it does what you need, it has good advice. You wake up in the morning, it says: here’s your daily report of how much progress your agents made overnight. Maybe it’s running a business for you — which I think is going to be a huge application of this technology. The democratization of entrepreneurship is absolutely coming. And it’ll say: here’s these problems, there’s this customer that’s upset, they want to talk to a real human, you should go talk to them. All of that is going to happen.

And then I think that the raising of the ceiling of ambition — of challenges humanity can solve — is also a next step for this technology, and we’re seeing the leading edges of it. The thing that I am just very excited to see is: if you remember AlphaGo move 37 — this move that no human ever would have come up with, that was creative, and it changed humanity’s understanding of the game — that is going to happen in every single domain. It will happen in science, in math, in physics and chemistry. It’s going to happen in material science, in biology, in healthcare, in drug discovery. But it may also even happen in literature, in poetry, in a bunch of other fields. It’s going to unlock human creative understanding and ideation in ways we can’t imagine right now.

Why do you think that hasn’t happened yet, given how strong you say the models are?

I think that there is an overhang of what the models are capable of and how people are using them. Our understanding of what is in these models is still emerging. So I think that even with no further progress, there’s still a massive shift that will happen. The economy being powered by compute and AI is still going to happen.

But I think there’s also something where what we’ve gotten very good at is training models on tasks that could be measured. So what we started with was math problems, programming problems where you have a perfect verifier. And a lot of what the progress has been in bringing us to more open-ended problems has been expanding the space of what can be evaluated. AI itself can really help with that — if AI is smart and understands things, you give it a rubric for how well a task goes. And for things like creative writing — is this a good poem? — that’s a much harder thing to grade. And so we’ve had less ability to teach the AI, for it to experience and try things out. But all of that is changing, and something that we have a lot of sight for.

AI Infrastructure Logic

As you shift really to these more agentic use cases, there’s been discussion about whether the bigger training runs really need to happen. If you get the model good enough, could you let it go out in the world, and then you can effectively get much of the uplift in areas that aren’t the pre-training, which is what these big data centers are needed for. What do you think about that argument?

I think it misses something very important for how the technology development goes, because it is absolutely the case that every single step of the model production pipeline multiplies, and so you want to improve all of them. And the thing that we see is: we improve the pre-training, it makes all the other steps much easier. And it makes sense, because it’s a model that is able to learn faster, a model that is already more capable to start when it’s trying out different ideas and learning from its own mistakes — that process is just faster, it needs to make fewer mistakes.

And so I think that the big shift has been from thinking of it as just training this cerebral system on its own, just making it bigger, to: it’s also about trying things out, it’s also about understanding how people are using it in the real world, and connecting that back into your training. But it doesn’t remove the value and the importance of continuing that research.

And the thing that I think has also shifted is we used to really just focus on the raw pre-training capability, but not think as much about the inference ability. And that’s been a big change over the past 24 months — to realize that it’s a balance. You can have this model that has all those great properties in the base, but then you really need it to be able to run at inference, because you need to do reinforcement learning, you need to serve it to the world. And that means you don’t necessarily go as big as you possibly could, because you also really think about all this downstream use, and you want the thing that has the best intelligence times that cost — to optimize those two things together.

Do you still need the Nvidia GPU if things move mostly to inference?

Read more

Iranian strikes have rendered two Amazon Web Services availability zones “hard down” in Dubai and Bahrain and the company expects them to be “unavailable for an extended period,” according to internal Amazon communication reviewed by Big Technology.

Within Amazon Web Services, the strikes have rendered so much damage that employees have been advised to deprioritize both regions.

“These two regions continue to be impaired, and services should not expect to be operating with normal levels of redundancy and resiliency,” an internal memo read. “We are actively working to free and reserve as much capacity as possible in the region for customers, and services should be scaled to the minimal footprint required to support customer migration.”

Reached for comment, an Amazon spokesperson pointed Big Technology to an Amazon blog post about the disruptions. “We continue to support affected customers, helping them to migrate to alternate AWS Regions, with a large number already successfully operating their applications from other parts of the world,” the post read. “As this situation evolves, and as we have advised before, we request those with workloads in the affected regions continue to migrate to other locations.”

With the war now nearing its sixth week, Iran has made Amazon infrastructure in the Gulf an economic target and is now eyeing its peers. Amazon’s Bahrain facilities have been hit multiple times, including a Wednesday strike that caused a fire. And its facilities in the UAE also sustained multiple hits. The IRGC is threatening multiple other U.S. tech giants, including Microsoft, Google, and Apple.

Amazons infrastructure in Bahrain and Dubai each have three ‘availability zones’ or clusters of compute. Both Bahrain and Dubai have a zones that are “hard down” and and “impaired but functioning.” per the internal communication.

“We do not have a timeline for when DXB and BAH will return to normal operations,” the internal post said.

This week on Big Technology Podcast: OpenAI President Greg Brockman: AI Self-Improvement, The Superapp Bet, Path To AGI, Scaling Compute

Greg Brockman is the President and co-founder of OpenAI. Brockman joins Big Technology to discuss OpenAI’s product strategy, the rise of its coming super app, and why he believes AI is entering a new takeoff phase. Tune in to hear Brockman explain OpenAI’s bet on the GPT reasoning model tree over video generation, what the “Spud” pre-training run means for upcoming models, and why he believes AGI is 70-80% achieved. We also cover the competitive landscape, the economics behind OpenAI’s $110 billion infrastructure bet, and public skepticism toward AI. Hit play for one of the most revealing conversations yet about where AI is headed and what it means for everyone.

You can listen on Apple Podcasts, Spotify, or your podcast app of choice

Thanks again for reading. Please share Big Technology if you like it!

Share

My book Always Day One digs into the tech giants’ inner workings, focusing on automation and culture. I’d be thrilled if you’d give it a read. You can find it here.

Join Big Technology’s Private Discord Server!

Where we’ll talk about this story, the latest in AI, the week’s podcast, and plenty more. You can sign up via the link below:

Read more

Senator Mark Warner seemed worried. Sitting in his office in Washington D.C., the three-term legislator from Virginia leaned forward, looked into his computer, and told me he didn’t believe the government, or anybody for that matter, was prepared for what might happen if AI reaches its potential.

“I don’t think the government’s ready. I don’t think society’s ready,” Warner said. “Short term, next three to five years, the economic disruption is going to be — I just think we are not ready at all.”

Warner is widely regarded as one of the most read-in U.S. lawmakers on technology issues, and I reached out to him after having a few recent unsettling conversations with AI-insiders who seem certain we’re under-appreciating AI’s potential near term impact. If there’s even a small chance that AI takes over swaths of knowledge work in the coming years, and that leads to job loss, the government should be right now working on plans to mitigate the downsides. Warner’s assessment wasn’t exactly reassuring.

“We’re at about 9% recent college graduate unemployment. I think that number will actually go to 30%,” Warner said. “To say government’s not ready would be an understatement.”

Multiple firms have already told Warner they’re using the technology to reduce their employee numbers, he said. One law firm is no longer taking on first year associates (“Going to take a pause,” Warner said, “and see how this works out before they even hire.”) Another midsized firm has reduced its back office team from 23 people to three. More “big brand name firms,” Warner said, are cutting internships in half or doing away with them altogether.

To be sure, the near-term predictions of mass job loss after ChatGPT’s debut have yet to come to fruition. Three years later, the U.S. unemployment rate is a not-disastrous 4.4%. But ever more capable AI makes the potential issues impossible to ignore.

In Washington, where legislators can barely keep the country’s airports open, part of what’s working against AI literacy is the complexity of the issues. “I think most members — and this is a human reaction — if you don’t get it and it seems too complicated, you want to try to punt on that,” Warner said.

Still, there are a few ongoing efforts that Warner and his colleagues are involved in that are, at least, a start. Warner and Senator Josh Hawley, for instance, recently introduced a bill that requires corporations and the federal government to report AI-driven job loss to the Bureau of Labor Statistics. And Warner and Senator Mike Rounds recently introduced the Economy of the Future Commission Act, which would generate policy suggestions to mitigate AI’s potential shock to the job market.

“They are, self-acknowledging here, small incremental steps when it very much could be the holy shit moment,” Warner said.

Asked if Congress is ready to meet the speed of AI’s progress with fast action of its own, Warner didn’t mince words. “I want to be more optimistic,” he said. “But I am terrified.”

Subscribe now

Is AI A Privacy Disaster? And How To Fight Back (sponsor)

Andy Yen is the CEO of Proton, the encrypted email and privacy software company. Yen joins Big Technology to discuss how AI is supercharging Big Tech’s ability to surveil us and what we can do about it. We dig into Proton’s ‘Born Private,’ which lets parents reserve a private email address for their kids for $1, and why the login-with-Google button is one of the most underappreciated privacy threats online. Hit play for a clear-eyed look at the hidden costs of Big Tech’s AI ambitions, and what a privacy-first alternative actually looks like.

Watch Now

What Else I’m Reading, Etc.

Apple will open Siri up to AI assistants from rival companies [Bloomberg]

ChatGPT’s controversial ‘Adult Mode’ has been shelved indefinitely [Engadget]

OpenAI’s advertising pilot is already above $100 million in annualized revenue [CNBC]

Citi says despite the S&P being down 6% on the year, there’s room to drop much further [CNBC]

This Week On Big Technology Podcast: Senator Mark Warner: Nobody’s Ready for What AI Could Do To Us

Senator Warner joins Big Technology to discuss whether Washington is prepared for the economic and societal disruptions of rapidly advancing AI. Tune in to hear why Warner believes recent college graduate unemployment could surge from 9% to 30% and why he’s more frightened than reassured about Congress’s ability to respond at speed. We also cover the Anthropic-Pentagon relationship, AI romantic relationships, data center opposition polling, and the ongoing battle over congressional stock trading. Hit play for a rare conversation with one of the few senators who actually understands what’s at stake.

You can listen on Apple Podcasts, Spotify, or your podcast app of choice

Thanks again for reading. Please share Big Technology if you like it!

Share

And hit that Like Button if you’re ready for more newsletters, at least.

My book Always Day One digs into the tech giants’ inner workings, focusing on automation and culture. I’d be thrilled if you’d give it a read. You can find it here.

Join Big Technology’s Private Discord Server!

Where we’ll talk about this story, the latest in AI, the week’s podcast, and plenty more. You can sign up via the link below:

Read more

On Tuesday, a federal judge in California will hear Anthropic’s argument to freeze the supply-chain risk designation the government slapped on it after it refused to grant the Pentagon the right to use its Claude models for domestic surveillance or autonomous warfare. The hearing is a key next step in Anthropic’s parallel lawsuits against the U.S. government, with one filed in California’s northern district and another in Washington, D.C.

Even as the case gains momentum, the Trump administration is already exploring new ways to force companies to scrap their AI safety and privacy protocols. Over the weekend, a new report from The Lever detailed a proposal that would require AI vendors to make their tech available for “any lawful government purpose” even if a company objects.

Ahead of this week’s hearing, filings from Anthropic, the government, and third parties help paint a picture of what’s at stake and the kind of legal arguments to expect. Here’s a brief primer:

The legal arguments:

Winning a preliminary injunction, which would freeze the government’s order to strip Anthropic out of all U.S. federal agencies, requires showing four things: Anthropic will have to show it’s likely to win the full case, that it would experience irreparable harm without the injunction, that it has more to lose than the government if the preliminary injunction is denied, and that the injunction would be in the public’s interest.

Anthropic’s case hinges on four main arguments:

• Its refusal to allow Claude for autonomous weapons and mass surveillance is protected by the First Amendment.

• The company’s Fifth Amendment protections were violated when the government blacklisted it without adequate notice or due process.

• Secretary of Defense Pete Hegseth exceeded his statutory authority by misusing a law reserved for foreign adversaries, by acting arbitrarily.

• President Trump’s government-wide directive goes beyond his authority, citing older cases from the 1950s and 1990s and more recent decisions like a court ruling Trump lacks authority to impose tariffs.

For the government to prevent an injunction, it’ll need to show Anthropic’s refusal to sign a contract term is commercial conduct and not protected speech. Lawyers will also argue the government has broad authority to choose its vendors and set contract terms as it sees fit. They’ll also need to show Anthropic’s privileged access to Claude creates a genuine national security risk, including the possibility it could alter the model’s behavior during active military operations.

Continued below: Anthropic’s chances of winning and what happens next…

Stop Phishing Cold. Secure Every Identity with Yubico. (sponsor)

Is your organization still relying on vulnerable, mobile-based MFA? It’s time to move beyond the security gaps of the past.

Yubico provides the gold standard in phishing-resistant authentication through device-bound passkeys. Our hardware-backed security keys offer a seamless, user-centric experience while eliminating the risk of account takeovers. Trusted by 19 of the top 20 global tech companies, Yubico empowers you to deploy a true zero-trust architecture at scale.

Download our latest guide, “Getting Started with Passwordless,” to learn how to future-proof your digital identity strategy today. Start leading with the highest level of assurance.

Download Now

Anthropic’s chances of winning:

Anthropic’s path for an early injunction is plausible. The company’s filings cite recent cases including Sen. Mark Kelly’s recent lawsuit against Hegseth, where a court ruled the government’s retaliation for his political speech was unconstitutional. (That case is still under appeal.)

Various filings suggest the law is on Anthropic’s side. Last week, a bipartisan group of nearly 150 retired federal and state judges filed an amicus brief supporting Anthropic in its parallel case in Washington D.C., arguing the law doesn’t provide a national security exception for the government to avoid judicial review.

Others point to recent decisions including the Supreme Court’s 2024 decision in Moody v. NetChoice, which ruled tech companies’ content decisions are protected by the First Amendment. One filing by a coalition of civil liberties groups suggests Anthropic should be given the same protections for decisions with its AI guardrails. Another filed by a group of retired military officers argues Hegseth exceeded his authority and that removing Anthropic now would hurt active military operations.

The path ahead:

A ruling from California’s Northern District could come within days or weeks. Based on the previous status briefing, U.S. Judge Rita Lin seemed to understand the case’s urgency and what’s at stake for both sides, even moving up the hearing after the U.S. government declined to commit to not taking any further action against Anthropic.

Even rivals have formally filed in support of Anthropic’s case. One amicus brief signed by dozens of AI experts argues Anthropic’s safety concerns are technically legitimate and not merely ideological. Signed by Google chief scientist Jeff Dean and others from both Google and OpenAI, the filing suggests the government’s supply-chain risk designation could have “serious ramifications for our industry” and expressed concerns about AI being misused for government surveillance.

“The mere existence of such a capability in government hands — even if never activated against a specific individual — changes the character of public life in a democracy,” the AI experts wrote.

The Intelligence Report

• An anti-AI march in SF took place over the weekend, with protestors giving speeches in front of Anthropic’s headquarters. Signs had slogans like “Don’t build skynet,” “Shut down OpenAI,” and “Stop The AI Race.”

• Jeff Bezos is reportedly in talks to raise $100 billion for a new AI manufacturing fund that would involve using AI to buy and transform old manufacturing firms.

• A rogue AI agent gave a Meta employee wrong technical advice that led to user data briefly being exposed.

• OpenAI published new details about its ad policies, offering more context for where ads in ChatGPT can appear, which advertisers are allowed, how sensitive topics are handled, and which types of conversations are still off-limits.

• Meta is keeping its metaverse alive after all, choosing to keep Horizon Worlds VR online just days after announcing plans to kill it.

• Speaking of the Metaverse, credit unions are setting up shop in Roblox to reach younger audiences.

• The White House released a new AI policy framework that includes pushing Congress to preempt many state AI laws. It also calls on Congress to make it easier to build AI data centers, to strengthen tools against AI scams, to improve copyright and digital-replica protections for creators and publishers and to expand AI training programs.

• Nearly a dozen major tech companies and retailers announced a pledge to share information about how scammers are using their platforms and services. It also calls for collaboration to improve AI-based tools for fraud detection and other solutions.

• Alex joined CNBC to discuss the latest in the Anthropic saga, and Bezos’s AI for Bezos’s new fund.

Will OpenAI’s Superapp Help It Refocus?

OpenAI is planning to simplify its product lineup with a new desktop “superapp” combining ChatGPT, Codex coding platform, and Atlas browser into a single destination. The new desktop app is part of a plan to streamline offerings and improve user experience for consumers, engineering and business customers.

The “super app” project will be led by OpenAI applications chief Fidji Simo, with president Greg Brockman helping revamp the product and teams working on it. At a recent all-hands meeting, Simo told employees that the company has become too distracted by different “side quests” while Anthropic gains momentum with enterprise and coding customers.

Call it a pivot or call it renewed focus, but OpenAI’s changes could also be a sign it’s worried about losing the lead. Simo and other top execs like CEO Sam Altman and chief research officer Mark Chen have spent recent weeks reviewing OpenAI’s entire product portfolio, according to The Wall Street Journal, which first reported the super app.

Anthropic already features its flagship Claude chatbot within the same desktop app as its tools for coding and computer use, but OpenAI still has separate apps for ChatGPT, Sora, Codex and its web browser. Meanwhile, OpenAI also has teams working on a secret hardware product being led by Jony Ive.

“Companies go through phases of exploration and phases of refocus; both are critical,” Simo wrote Thursday on X. “But when new bets start to work, like we’re seeing now with Codex, it’s very important to double down on them and avoid distractions. Really glad we’re seizing this moment.”

OpenAI also faces growing backlash after it rushed to win a contract with the defense department and replace Anthropic. According to the grassroots website QuitGPT, more than 4 million people have vowed to delete ChatGPT, up from 1.5 million on March 4 a few days after landing the new defense contract.

A New AI Film For Non-AI Experts

On Friday, a new AI documentary will arrive in theaters and give mainstream audiences their first real introduction to the debate that has largely consumed Silicon Valley and Washington, D.C.

“The AI Doc: Or How I Became an Apocaloptimist” explores AI’s potential promise and peril through the perspective of co-director Daniel Roher, who asks dozens of AI experts earnest questions about how AI is reshaping the world his son is about to be born into.

The result is a film that spans the full spectrum of AI opinion — doomers, skeptics, realists, and accelerationists — without stacking the deck. It’s already been screened at Sundance and SXSW, aims to help people get a better understanding of AI and have proactive awareness for what’s at stake. In an interview this month, the creators noted it’s disappointing how unprepared AI execs seem for dealing with the potential risks.

“In the film we go to the CEOs, because they’re the guys building this,” co-director Charlie Tyrell said while in Austin for SXSW. “...As you know in the film, you go there and there is no plan. So that puts the onus on the users, the non-technologists, to ask what we do.”

The cast is a who’s who of the AI debate across the full spectrum of opinion. It has researchers and ethicists sounding alarms about AI, with people like Timnit Gebru, Deborah Raji, Yoshua Bengio, Karen Hao, and Emily Bender. It also includes AI execs Sam Altman, Anthropic’s Dario and Daniela Amodei, Demis Hassabis, and Reid Hoffman.

One prominent voice in “The AI Doc” is Tristan Harris, co-founder of the Center for Humane Technology, who also mentioned the film in his SXSW talk earlier this month. Harris argued there’s still a lack of common knowledge about AI, but that the film helps “create global clarity about why AI is actually dangerous and what the current path looks like.”

Big Technology Friday Edition: OpenAI’s Superapp Ambitions, Jensen on Jobs, Bezos’s $100 Billion Automation Fund

Ranjan Roy from Margins is back for our weekly discussion of the latest tech news. We cover: 1) OpenAI leadership says no more side quests 2) The company is focusing on enterprise and coding 3) Does this mean consumer AI is dead? 4) OpenAI’s new focus era 5) Why OpenAI is building a Superapp 6) OpenAI partners with the consultants 7) Most first time AI buyers are choosing Anthropic 8) Nvidia CEO Jensen Huang says those who use AI to cut jobs lack imagination 9) The Metaverse is dead, or is it? 10) Jeff Bezos is raising $100 billion to automate industrial work 11) Do you dry chat?

You can listen on Apple Podcasts, Spotify, or your podcast app of choice

Thanks again for reading. Please share Big Technology if you like it!

Share

Join Big Technology’s Private Discord Server!

We’ll talk about this story, the latest in AI, the week’s podcast, and plenty more in our private Discord server. You can sign up via the link below:

Read more

As Nvidia CEO Jensen Huang takes the stage at his company’s flagship GTC event on Monday, the stakes will be especially high. Nvidia’s annual conference is always a major moment for the company’s products and roadmap, but this year GTC will look more like a legitimacy test for the next phase of AI buildout.

Nvidia’s continued expansion depends on high le…

Read more

When Andrew Ross Sorkin released his book on the 1929 stock market crash last year, the prevailing worry was that the debt-reliant AI buildout might end in disaster. Now, six months later, the market is freaking out about AI’s potential to succeed. Every Anthropic blog post seems to cause a mini-meltdown in software stocks and unease about AI’s potential to replace jobs is building fast.

So, this was the right moment to check in with Sorkin, a CNBC anchor and New York Times columnist, about whether today’s set of concerns about AI’s success are warranted, and whether they could hit the market hard if realized. In an hour-long conversation at the New York Stock Exchange this week, we covered the labor and capital side of AI, how the AI bet might impact private credit, prediction markets, the SpaceX IPO, and plenty of other ground.

You can read our full conversation, edited lightly for length and clarity, below. Catch it next Wednesday on Big Technology Podcast, which you can subscribe to on Apple Podcasts, Spotify, or your podcast app of choice.

Alex Kantrowitz: Hi Andrew, Market crashes are typically caused when debt goes bad. So how do you evaluate the current concern that AI works and everything is disrupted?

Andrew Ross Sorkin: I’m often asked, is there a modern day way to get to 1929? And what people are really asking is, is there a modern day way to get to 1932 — which is 25% unemployment in America?

I always think the answer is actually less of a market crash and more AI. If you ever wanted to think about what would this country look like with 25% unemployment, how would you get there? And I think the answer is, potentially, if AI is as successful as I think we all hopefully want it to be — to the extent you believe these valuations are real — all of the math behind it, the only way that really works to some degree, is to create extraordinary productivity. And what does productivity mean? Well, it means a lot of growth at a lot less cost. How do you take out that cost? Well, we’re both looking at each other, and that’s pretty much — we are the cost.

The robot employee is going to be a thing.

It’ll definitely be a thing. My kids talk about it being a thing. I’ve got 15-year-old boys, and we talk about what they’re going to do. But then we talk about, literally, are we going to have a robot in our house in five years from now, and what are all the things that the robot is going to do?

I am skeptical that AI is going to cause this wave of mass unemployment. If it’s able to do the jobs of, let’s say, 20% of the workforce — wouldn’t you anticipate a production boom that would come along with that and help grow the economy?

That, to me, is the question. It’s not, are there more things to do — it’s who’s going to have the money to do those things?

I don’t think that we have to have mass unemployment forever, either. I think it’s possible that there could be a painful transition period. And historically, when we’ve gone through these technological revolutions, there have been painful transition periods. And so if you’re a young person today, doing the job that, frankly, increasingly it appears that a Claude or ChatGPT could even do — whether it’s research or putting together a model or being a paralegal, or name your role — you say to yourself, if you’re running one of these firms that historically hired kids out of college to do that, are you going to still hire those kids to do that? Is there a higher order kind of work that they can do for you now that you can get this work done by the AI? I think these are the real questions.

And then there’s going to be an economic one, which is, tokens are not free. AI is not free. But how much cheaper is it ultimately going to be than a human?

Our conversation continues below…

A MESSAGE FROM OUR SPONSOR

See autonomous AI in action

Feeling forced to choose between AI innovation and control? Discover how
agentic AI can deliver both — completing requests end to end, securely and with enterprise-level governance. The result: faster outcomes, stronger oversight and more capacity for high-impact work. Watch the demo.

When you look at the statistics, though, it’s very interesting — AI is more sophisticated in coding than anything else, but we’re definitely not seeing a wave of layoffs of software engineers now.

You don’t think that in two years from now, just in terms of the magnitude step change in how good the technology is going to be, that it’s not going to get that much better? I agree, if this is the level, I’ll bet with you. But if you believe in the technology improving — which invariably it has to — and by the way, if it doesn’t, then we’re in a whole other different world. Then we’re not talking about what happens in success. We’re talking about what happens in failure, because then we really will have a bubble.

But if it doesn’t improve the way I think the model makers — policymakers, investors — want it to, I just don’t see how we’re going to be sitting around doing our own programming. I just don’t see it. By the way, I write my own articles today. I wrote this book without AI. This took eight years, unfortunately. But five years from now, do you really think that people are going to write books even by themselves? I assume you will be co-authoring a book with AI. I imagine people will be writing articles, at minimum, with AI — if AI is not doing it entirely, in which case, then there’s a question about what is the role of the human in all of this.

The thing that I wonder about is whether businesses in the economy will be content with what they’re doing today or just go after their roadmap in a way they’ve never been able to before, because they’ve been constrained by labor.

That generally assumes that then you have to be able to massively upsize the size of the pie, right? This is growth for everybody. But there is some of this that is a zero sum game. It is not like the pie can just grow exponentially. I know there are people who believe that it could, but invariably, at least historically, it hasn’t.

So yes, I imagine there’ll be people who do even more. But I would also say — I walked in here and I described you as an independent media titan, right, in this role that you’re in. You’ve got a sort of satellite group of people who you use and work around you. And maybe over time, you hire some more people here and there, but maybe you wouldn’t in the future, and maybe the sort of network that you’ll have will be your agents that will do this for you. Well, if you don’t have those people doing that work for you, what are those people doing?

I myself am a little bit of a small business. As I was promoting my book, I ended up having to hire a couple of different people — social media people, this and that. And I was sent a contract which I needed to fill out, and typically I would have sent it off to a lawyer who would have charged me, I don’t know, a couple hundred dollars, maybe a thousand. And what did I do? I took the contract. I put it in ChatGPT. I said, tell me everything that’s good and bad in this contract. It spotted the things that I already saw and spotted some others. It then says, do you want me to redline the contract and mark it up? I say sure. It then says, would you like me to write a cover letter back? I say sure. I make some changes to it, I make sure it hasn’t hallucinated, and it’s off.

But it means that if every other small business owner operated the way I did in that moment, all of the lawyers who do business for small businesses for things that are not that complicated — these were contracts with very little at stake — I think people wouldn’t probably use lawyers for those things.

Now, then you say, well, maybe then the lawyers are going to have to figure out, can they be doing work that’s even higher grade? What is that higher grade work? We already have that higher grade work. It’s what big corporations look to them to do for mergers and acquisitions and other things. But the small business lawyer typically hasn’t been doing that work in the past.

That answer goes back to the question — is there going to be a limit on growth? If you were to take the example that you just gave, where now you’re able to go in and negotiate this contract with AI. Well, all of a sudden you get time back and you can work on improving your book, writing another article for The Times, researching a segment for the next day’s Squawk Box. You become that much more productive, right?

But I’m now not employing more people. So I do believe — talk about inequality — I believe that the wealth and the great riches are going to go both to the model makers, some of the big tech companies, and probably the folks who already have had success, because they will be at the top of these food chains, and instead of hiring more people, they’re going to hire more agents.

There are places though where I could definitely see some real disruption, like accounting. After watching Claude Code go out and build computer programs autonomously by following prescribed rules, it will probably be able to grab the code for a certain region and balance the books.

There’s much of that throughout our entire economy. Take journalism, for example. You’ve got a football game on over the weekend. You’ve got the scores. You know what happened during the game. The value proposition for a journalist to be watching that game and ultimately be reporting on it is their ability to analyze the game, potentially as a columnist, to be able to explain what happened in an entertaining way, but maybe have some insight into the player, maybe they had a relationship with the player, knew the player had been in the exact same position five years ago, and they have a whole bunch of different stats and all of these things.

Some of that — not all, but some of that — I imagine AI in the future will be able to replicate. There are other parts AI can’t do: walk into the locker room and start to interview the athlete, or figure out what the coach did two days earlier and try to get somebody to tell them something off the record or behind the scenes. And that’ll become a sort of higher order value proposition. But there’s a lot of things that are part of our daily life that I imagine will get automated.

This has existed for a long time though. Narrative Science is a company that’s been able to take the box score and turn it into a story. But we still love watching ESPN and watching live sports.

I don’t think we’re going to stop watching live sports. The question is, are we going to read about them in the same way? The service you described is looking at the delta between the scores, and then is able to describe what happened in a particular way. What happens when the AI can actually watch the game itself?

You know, if Sam Altman and Johnny Ive have their way, we will have something sitting on this desk probably in six months from now that may be watching what we’re doing all the time, right? In which case it’ll have a persistent memory about all of our interactions. And maybe if we’re watching the game, it’ll be able to report on the game itself.

Nothing will beat the reporter going into the locker room and speaking to the losers and winners. The people matter.

Google’s NotebookLM, for instance, makes amazing AI podcasts on any topic you could want. There was one podcast that hit the top 30 trending on Spotify that was entirely AI-created, about the Epstein files.

But by and large, we have the ability to create these shows, and we would still rather see two human beings have that discussion….

You’re freaked out by that Epstein podcast??

No — you’re actually giving me an idea. He just pointed, for those listening, he pointed to the book, 1929. I’m thinking maybe NotebookLM could put together an awesome podcast on 1929.

So there’s the labor risk, but there’s also a potential hollowing out of the software industry, and maybe other industries. It seems like we have these colliding forces where either the bet on AI needs to go bust or there’s going to be some serious consequences for everybody else.

So maybe I’ll be on the other side of this one. I’m not sure that software is dead just yet, because I think to myself, sure we can build our own model for whatever app we want to build for our company. But at the same time, a lot of these software companies already have an install base. They have some data. And on top of that, they probably should be — and are — using AI too to build their apps and their software. So I would imagine, unless you think that everything is just going to be built either by individuals and companies, or we’re all hiring Accenture to come in and just build custom software for everybody, I would think that some of these software companies will actually continue to have success.

The companies that are not the AI chatbot makers tend to think that there’s going to be a set of chatbots — there’s going to be a bot you use when you want to have a conversation like a ChatGPT style one, but one where you shop, and one where you’re doing your enterprise stuff and learning about what’s going on inside the company. I just see it consolidating.

I’m with you on that. I think we are all going to have one bot. Now the question is, does that bot do a sort of secret or quiet handoff to another bot, and it’s seamless to us, and we don’t even know — meaning our interaction is going to be just with our bot. And it’s not just that there’s one big model that’s going to do everything, but it’s that our bot says, oh, Sorkin wants to shop right now, there’s a specialty shopping bot over here, I’m just going to hand them off to that bot, and I’ll tell the bot everything I know about Sorkin so that it feels as if it’s the exact same bot working with them.

I think that’s probably going to happen, but the interesting thing is — I think this is the bet among many tech companies right now — they don’t fully know, we don’t know how good the core bot gets. So the reason why you would hand off to a shopping bot is because it’s difficult to build shopping capabilities into a general purpose bot right now. It takes a lot of planning, specialty knowledge. But once you get a bot that can do everything in this AGI way, maybe that need to specialize goes away, or maybe it can teach itself.

But I do think there’s going to be one interface that you will interact with.

Likely talking rather than even typing. By the way, I’m talking out constantly to my phone in a way that I wasn’t six months ago. Do you do that?

Definitely all the time. We have three Alexa Pluses in the house, and we’re talking to Alexa Plus all the time.

Are you typing your newsletter?

I’m typing it. I still believe in writing unassisted by AI, because to me, that’s the only way to think. Type it out.

What about texting? So my wife sends me a text saying, are you late? Which is typically what I am. And then it used to be that I’d write ‘yes’ or ‘I’m 20 minutes behind.’ And now I just say, yeah, I’m running late, sorry, I’ll be there shortly. I just feel like I’m doing that all the time now, a completely different way.

Now, you’ve said that if the AI companies can’t make good on all the money that’s been invested — and there’s a lot of debt there — then we could see some form of crash.

Has the fact that they’re starting to make real revenue made you feel a little bit better?

I feel better about it in two contexts. One is that they’re making more money. And two is that, I think when you really dig under the covers of a lot of these commitments that these companies have made to others — meaning data centers that are going to be built on the back end of these things, or even some of the investments we heard being made early on, saying they were putting in a hundred billion — when you realize that everything was tranched, those tranches make it safer, because it doesn’t mean that you’re going to go out and spend a hundred billion tomorrow even though you don’t have it today. So I do think that we’re hopefully in a bit of a better situation in that regard.

I still don’t think we’ve taken full account, though, of two component parts of this. What happens in great success from a technology perspective, in terms of the efficiency of these models? Could we ever get to a point where you actually don’t need all of these data centers, where a lot of the compute moves to the edge, and then all of a sudden the economics of that get upended? And then the other side is, could the depreciation schedule of these chips either be way shorter or way longer than we think? So I think there’s still a whole bunch of pieces of the puzzle that we haven’t figured out.

Yeah, that’s kind of one of the hypotheses we’ve been playing with on the show now — Did Apple just do it right? Where Apple becomes the infrastructure for AI, where it happens on your phone and on the Mac Mini, and the models become so efficient or purpose-built that you don’t need the data centers.

It may very well be that that’s where this lands. Having said that, I would imagine that this should be the greatest opportunity for Google and Alphabet to truly take share. Because Gemini, if built the way I think it should be built, should be able to move you around their phone in shocking ways. It should be able to move into any app, control any app, do everything throughout the phone.

I can’t imagine that Apple, which really has made its name around privacy and controls and a sort of walled garden, even under this new deal that they’re going to have with Google using Gemini as sort of their Siri, are going to let whatever that is go super deep throughout the entire phone. And so I would imagine that’s why I’ve always thought this should be Google’s time.

Private Credit

Let’s talk about debt. One of the interesting things that you’ve said repeatedly as you’ve discussed your book, is that if there’s too much leverage, that’s where you have the problems.

AI firms have been taking on a bunch of this debt for the infrastructure buildout. And I just find it astounding that we don’t know where the debt is in the system, because so much of it is happening in private credit.

I mean, look — we’re seeing right now a lot of hand wringing around private credit. A lot of questions about what that whole market is going to look like. And if that market seizes up, what does that do to lending in other parts of the economy? What does it do to real estate? You could see how it could have a big impact.

I will say one thing we haven’t really wrestled with is that the private credit business really can’t fall apart before, frankly, the private equity industry falls apart. Meaning, if you think about it, who has taken on the biggest loans from the private credit space? It’s oftentimes the private equity players, or some of these tech players. So the private credit people can’t lose money unless the other people lose money first.

And we really haven’t grappled with that, and that’s in large part because so many businesses today are private companies, and the marks, the valuations, are not in the public market. It’s not a day-to-day operation. And there is a whole universe that I’d put in the category of mark-to-make-believe, and the incentive system is such that you want that to go on as long as possible. Because if you are the equity owner, you don’t want to mark your position down, because then you can’t raise new money. And if you’re the private credit firm, you also don’t want them to mark it down, because it’s going to impact your own value. So you can see that it’s a cascading effect.

There are a whole lot of people who are holding on as tight as they can, hoping they can bare-knuckle this thing. Maybe we can get to a place where Kevin Warsh gets in the seat and we get some lower interest rates, and things kind of ease off, and maybe the world gets a little easier.

For those who aren’t fully into this — can you just briefly explain what this private credit, private equity thing is and why it’s not happening in the big banks?

Read more

SXSW 2026 kicks off on Thursday and AI news should be pouring out of Austin.

This year’s “Tech & AI” track features about two dozen sessions, making up about a third of the festival’s total featured talks. AI-related sessions will feature a mix of researchers, media executives, and technologists including Spotify co-CEO Gustav Söderström, Cloudflare co-founder Matthew Prince, and Apple Fellow Phil Schiller.

The frontier labs and their critics will be represented across various panels and workshops, including a talk with OpenAI execs about infrastructure projects like Stargate and a fireside with Tristan Harris, co-founder of the Center for Humane Technology, about run-away AI.

One of the quirkier events on the tech calendar, SXSW’s assortment of CEOs, ethicists, futurists, and filmmakers should provide us with different perspectives beyond the whims of the usual suspects.

A few AI-related themes to expect from SXSW Interactive 2026:

• AI moves from models to deployment: The conversation is shifting away from LLM hype toward the rest of the stack: infrastructure, agents, enterprise applications, and more for consumer businesses, creators, media companies, etc.

• Living with AI’s consequences: Instead of debating if AI is here to stay, this year will feature more talks about ways people are dealing with the new reality. Many sessions focus on governance, guardrails, intellectual property, education, and the long-term cognitive effects of AI.

• The post-search internet: Multiple panels explore how AI agents and generative search could reshape news, web traffic, and media economics. Speakers include BuzzFeed’s Jonah Peretti, Mark Cuban (“Can Media Survive AI?”) and a session on “The Internet After Search” featuring Cloudflare co-founder Matthew Prince.

A few predictions for who might make news at SXSW 2026

There will likely be plenty of news throughout the next week, from AI companies, other tech companies,and consumer brands using tech in new products. Announcements will likely come from hardware, software and everything in between. For example, Rivian will reveal its new R2 vehicle.

One company to keep an eye on this week is Nothing, whose CEO Carl Pei will be in Austin for an SXSW featured talk about shifts for things like AI hardware. Even if Nothing doesn’t announce anything new, SXSW could offer a glimpse of what the next wave of AI-powered wearables might look like after the splashier launches at CES.

More on the SXSW calendar below….

Launch fast. Design beautifully. Build your startup on Framer — free for your first year. (sponsor)

First impressions matter. With Framer, early-stage founders can launch a beautiful, production-ready site in hours. No dev team, no hassle. Join hundreds of YC-backed startups who launched here and never looked back.

Pre-seed and seed-stage startups new to Framer can enjoy:

• One year free: Save $360 with a full year of Framer Pro, free for early-stage startups.

• No code, no delays: Launch a polished site in hours, not weeks, without hiring developers.

• Built to grow: Scale your site from MVP to full product with CMS, analytics, and AI localization.

Join YC-backed founders: Hundreds of top startups are already building on Framer.

Apply to claim your free year →

Learn More

Here are the top AI talks happening during the first half of SXSW 2026.

March 12:

• Rivian will officially reveal the R2 at its Rivian Electric Roadhouse, which will be open throughout the festival. On March 13, the company will host another event called “Are you Faster Than a Robot?”

March 13:

• One of the featured sessions will be a talk with Gustav Söderström, Co-Chief Executive Officer Spotify, will speak on stage about the “Past, Present And Future Of Delivering Creativity To The World” with singer/songwriter Lainey Wilson and Ohalo CEO David Friedberg.

• Steven Spielberg will join SXSW for a keynote about the future of movies, moviegoing and his upcoming film “Disclosure Day.”

• In a session titled “How We Could Lose Control: Avoiding the Paths to Runaway AI,” Tristan Harris and physicist Anthony Aguirre will discuss how humanity could lose control of AI systems and the steps needed to prevent it.

March 14:

• Amy Webb, CEO of the Future Today Strategy Group, will deliver her annual Emerging Tech Trends report. This year’s theme is “creative destruction.”

• Execs from OpenAI and the San Antonio Spurs discuss how the NBA team is using AI to personalize experiences for fans, applying human-centered design and improving internal decision-making.

• Cloudflare co-founder & CEO Matthew Prince and Manueto Ventures CEO Stephanie Mehta will talk during a session about “The Internet After Search.”

• Experts from Microsoft, the Allen Institute and the University of Washington will discuss AI and future scientific discovery.

March 15:

• Tech journalist Karen Hao, computer scientist Timnit Gebru, and MacArthur Foundation president John Palfrey discuss how to build AI systems and governance frameworks that prioritize humanity instead of just Silicon Valley.

• David Rogier, MasterClass founder and CEO, will discuss how AI could reshape storytelling through personalized narratives and offer ways to challenge assumptions about creativity and culture.

The Intelligence Report

• Anthropic and the Pentagon continued their dispute, with the government officially designating the startup as a “supply chain risk,” forcing federal agencies like the State department, Treasury and others to pause their use of Claude.

• OpenAI, which faces backlash over its Pentagon agreement, is reportedly in talks for a NATO contract related to unclassified networks.

• Meta signed a new AI data licensing deal with News Corp with a contract reportedly worth up to $50 million a year.

• OpenAI debuted its latest model, GPT-5.4, which the company claims cuts hallucination rates by 26.8% when using the web and by 19.7% when using internal sources.

• Google released Gemini 3.1 Flash-Lite, which the company says will help developers handle high volume tasks at scale with lower costs.

• The White House and seven major tech AI companies agreed to a new “ratepayer protection pledge,” with Microsoft, Meta, OpenAI, Oracle and others agreeing not to pass on costs to consumers from AI data centers.

• Meta faces a new class-action lawsuit over its Ray-Ban Meta AI smart glasses over privacy concerns that involve allegations of Meta workers watching users’ sensitive content.

• Apple debuted its new M5 chips for the MacBook Pro and MacBook Air, which promise to offer better on-device AI performance.

• Oregon lawmakers passed a new AI bill to improve chatbot safety rules for kids in the state. The bill now goes to the governor for approval and could be the first major AI chatbot safety bill in a state to pass this year.

Oracle’s layoffs: A preview of more to come?

In the past month, several high-profile companies have trimmed headcount as their AI investment ramps. Just last week, Oracle became the latest to announce thousands of cuts in order to free up money for expanding its data centers.

Before that, Block announced plans to reduce its workforce by 40% in order to rebuild the fintech company “as an intelligence.” And in January, Amazon said it would cut 16,000 jobs.

This may just be the start of a new tech layoff season. For some, the AI narrative makes for a convenient explanation for layoffs driven by more familiar forces like cost discipline or economic uncertainty. For others, changes might actually reflect genuine attempts to redesign organizations around new tools and workflows.

Tech companies have the impossible task of planning for an uncertain future, all while costs are going up. Often, it’s human employees at the receiving end of this trend. And these types of cuts can be contagious. Prepare for more on the way.

ICYMI: Michael Pollan on AI and consciousness

Michael Pollan, author of a newly released book “A World Appears,” recently joined Big Technology Podcast to talk about why he doesn’t think AI will ever be conscious. After decades of exploring the minds of both humans and plants, he’s much more skeptical that machines are able to achieve consciousness in the same way humans do. In fact, he thinks using the human brain as a metaphor for computers has serious flaws.

Unlike computers, human brains don’t have the same sort of “hard separation,” with the former operating in analog mode instead of digital. Pollan also thinks feelings are central to consciousness, and that’s still something that’s impossible for computers to experience.

The way Pollan sees it, having a conversation with a bot or any computer reduces or simplifies the very notion of what a conversation is: “You’re leaving out what’s going on between us right now, which is acknowledgement, skepticism, body language, all the subtleties of human conversation are stripped away.”

“The paradigm case is the emoji [and] accepting the emoji as a substitute for emotion,” Pollan said. “We have to be careful about that when we simplify these phenomena like machine consciousness… What are we doing to the word ‘relationship’ when we count that?”

You can listen on Apple Podcasts, Spotify, or your podcast app of choice

Share

Join Big Technology’s Private Discord Server!

Where we’ll talk about this story, the latest in AI, the week’s podcast, and plenty more. You can sign up via the link below:

Read more

This week, I was surprised to learn that the world’s leading AI labs have granted themselves free rein to train on our conversations. I’ve since fixed the problem for my account, and you might want to as well.

Amazon, Anthropic, Google, OpenAI, Meta, and Microsoft have all built default settings that allow them to train on anything you input into the chatbot window, from medical records to open-hearted confessions. Unless you toggle the setting off, you’ve granted them the right to access all your AI interactions.

If you’re mostly using the bots for rudimentary work, that’s probably fine. But if you’re inputting financial, medical, or other personal information (I’m guilty of all the above), then it’s less advisable.

“You’re opted-in by default,” Dr. Jennifer King, privacy and data policy fellow at Stanford’s Institute for Human-Centered Artificial Intelligence, told me. “They are collecting all of your conversations.”

Dr. King is the lead author of a viral paper that examined these companies’ data collection processes last year. The paper, called User Privacy and Large Language Models, highlights a privacy issue that will only become more urgent as people trust increasingly capable chatbots with more sensitive documents.

And given that these AI research outfits had exhausted almost all available data on the internet (and elsewhere), new data coming in via our conversations with the chatbots is particularly precious. “It’s very valuable“ said King. “The research today shows that if you keep retraining on AI-generated content, you end up with model collapse.”

The AI model builders do install guardrails to ensure personal information isn’t spit out by chatbots. And many strip identifying information out of training sets. But with the amount of data involved, it’s a risky process to trust.

How to Opt-Out

The chatbots’ settings are somewhat hard to find and discern. So in the spirit of full disclosure, I’ll lay it out here, at least for ChatGPT, Claude, and Gemini:

ChatGPT

Within the Data Controls section in ChatGPT’s settings, the “Improve the model for everyone” opts your conversations in for training. Turn that off and the company no longer has your permission to train on your chats.

Claude

Inside Claude’s Privacy section, the “Help Improve Claude” setting can be toggled off to remove your conversations from training.

Gemini

And for Gemini, head to the Activity section within settings and turn that off to prevent training:

The squishy opt-out language is “obscuring what they’re really doing by appealing to your social good, essentially,” Dr. King said. ”It frames it as a trade-off, that you’re going to make this thing worse if you don’t comply.”

Now, at least you have the information. Go do with it what you will.

What Else I’m Reading, Etc.

Anthropic and the Pentagon are back in talks [CNBC]

OpenAI’s new GPT 5.4 model can natively use computers [TechCrunch]

Yes, we can tell when you’re writing with AI [WSJ]

Cluely CEO admits he lied about $7M revenue figure [TechCrunch]

The McDonald’s CEO couldn’t bring himself to scarf his burger [New York Times]

Survivors of the Tahoe backcountry avalanche share what happened [New York Times]

This Week On Big Technology Podcast: Pentagon Insider: What’s Next For Anthropic and The Department of War — With Michael Horowitz

Michael Horowitz is the former deputy assistant secretary of defense for force development and emerging capabilities at the Department of Defense, and currently a professor at the University of Pennsylvania. Horowitz joins Big Technology to discuss the Anthropic–Pentagon rupture and what it signals about how the U.S. government wants to use frontier AI. Tune in to hear his inside view on how models like Claude actually get deployed in defense workflows, why a contract fight over “mass surveillance” language escalated, and what the trust breakdown says about the future of AI partnerships with the state. We also cover autonomous weapon systems vs. “fully autonomous weapons,” what today’s AI can and can’t do on the battlefield, and how AI is likely to reshape warfare over time. Hit play for a clear-eyed look at where Silicon Valley and the national security establishment collide—and what happens next.

You can listen on Apple Podcasts, Spotify, or your podcast app of choice

Thanks again for reading. Please share Big Technology if you like it!

Share

And hit that Like Button it won’t collect your deepest darkest thoughts, I promise

My book Always Day One digs into the tech giants’ inner workings, focusing on automation and culture. I’d be thrilled if you’d give it a read. You can find it here.

Join Big Technology’s Private Discord Server!

Where we’ll talk about this story, the latest in AI, the week’s podcast, and plenty more. You can sign up via the link below:

Read more

A decade ago last month, Apple CEO Tim Cook wrote a “Message to Our Customers” explaining why the company refused to build an encryption backdoor after the FBI demanded access to an iPhone during its investigation of the 2015 San Bernardino shooting.

In the blog post, published February 2016, Cook warned that forcing Apple to weaken its security could create a slippery slope that might make it easier to someday force it to build surveillance software and “undermine the very freedoms and liberty our government is meant to protect.” The FBI eventually found another way into the iPhone, but the fight helped solidify Apple’s reputation as a privacy-focused company and build customer trust that’s lasted much longer than its brief legal battle.

A decade later, it’s Anthropic instead of Apple on the hot seat. And today, Anthropic CEO Dario Amodei’s refusal to allow the Pentagon to use his AI for domestic surveillance and autonomous warfare (at least contractually) could add a similar halo to Anthropic’s brand. But it’s much less straightforward for Anthropic than it was for Apple, as the rising AI lab faces consequences that are coming into focus this week.

Anthropic has, by all accounts, had a hellish few days, enduring repercussions Apple never did in its face-off with the FBI. After being publicly attacked by Pentagon officials, the company was targeted in a social media post by President Donald Trump, who deemed it a supply chain risk and ordered every federal agency to stop using its technology. Already, Treasury Secretary Scott Bessent has said the Treasury Department “is terminating all use of Anthropic products.”

Within six months, government suppliers must also stop using Anthropic for their government work. Some are already considering going further than that, preemptively, to stay in the U.S. government’s good graces. Should the government go through with its Claude ban (still an open question) it would cost the Anthropic hundreds of millions, if not billions of dollars.

Labeling Anthropic as a supply chain risk is “attempted corporate murder,” as Trump’s former AI advisor Dean Ball put it.

Share

Claude Rises

Anthropic will be around for a long time though, and if it was looking to use this moment to assert its values, as Apple did a decade ago, that appears to be working. Soon after the company bucked the Pentagon, supporters chalked “God Loves Anthropic” in front of its San Francisco headquarters, PR and legal experts took its side, Katy Perry posted a screen shot of her signing up for Claude Pro, and AI professionals, broadly, came to its defense.

“I’ve honestly never had a high opinion of Anthropic before,” Chris Paxton, AI innovation lead at Agility Robotics wrote on X. “But I think I’ve come around to believing that these folks actually believe what they are saying, and have gained a lot of respect for them.”

Over the weekend, Claude climbed to the top spot in the Apple App Store for the first time, jumping from No. 42 before its first Super Bowl ad a few weeks ago. Claude’s total free users, meanwhile, have grown more than 60% since January while paid subscribers have more than doubled this year, according to the company. Another chart from AppMagic shows Claude’s total global downloads in a single day reaching 500,000 for February 28.

Claude’s website also now includes a page showing users how to import memory to Claude from other AI providers, a clear indication its looking to capitalize on this moment.

OpenAI Swoops In

As leading models converge on capabilities, the AI sector will differentiate on brand. And on Friday, when OpenAI agreed to a similar deal to the one Anthropic wanted, but ultimately walked away from, it set a line of delineation between the two companies.

OpenAI stands to make hundreds of millions of dollars from the deal, but already, protests are forming in response. Tomorrow, “QuitGPT” organizers are holding a “No Killer Robots, No AI Surveillance” protest outside OpenAI’s headquarters. “​Last Thursday, Anthropic refused to let the Department of War sidestep its basic safety guidelines,” its event page says. “A day later, OpenAI swooped in and agreed to let the Pentagon use its AI for killer robots and mass surveillance of American citizens. ​Protest this CORRUPT deal at OpenAI’s front door.”

Soon, we’ll learn whether Amodei’s dealings with the Pentagon will propel Anthropic forward as it competes with OpenAI, or whether the damage done by Trump’s order sets it back decades. Tim Cook’s experience may be instructive, but following the trend of most AI news lately, this situation is unlike those that came before.

Subscribe now

Launch fast. Design beautifully. Build your startup on Framer — free for your first year. (sponsor)

First impressions matter. With Framer, early-stage founders can launch a beautiful, production-ready site in hours. No dev team, no hassle. Join hundreds of YC-backed startups who launched here and never looked back.

Pre-seed and seed-stage startups new to Framer can enjoy:

• One year free: Save $360 with a full year of Framer Pro, free for early-stage startups.

• No code, no delays: Launch a polished site in hours, not weeks, without hiring developers.

• Built to grow: Scale your site from MVP to full product with CMS, analytics, and AI localization.

Join YC-backed founders: Hundreds of top startups are already building on Framer.

Apply to claim your free year →

Learn More

The Intelligence Report

• OpenAI said ChatGPT passed the 900 million weekly user mark and now has 50 million paying subscribers.

• Block, the parent company of Square and Cash App, announced plans to lay off 40% of its workforce as part of a plan to remake the company for the AI era.

• Two founding members of Mira Murati’s Thinking Machines Lab left Meta, marking the latest in a string of departures from the startup founded by OpenAI’s former CTO.

• AMD announced a new deal with Meta, which has agreed to buy billions of dollars worth of chips to develop AI tech and power new data centers.

• OpenAI announced a new deal with four top consultancies to create a new Frontier Alliance, which includes OpenAI engineers having forward-deployed engineers working alongside to develop AI solutions.

• The AI search startup Profound raised another $96 million at a $1 billion valuation to help brands figure out how to be visible to LLMs.

• Nvidia announced record quarterly results, which still failed to move the needle for Wall Street. It also announced plans to invest $30 million in OpenAI’s new $110 billion mega-round, which the startup announced also had participation from Amazon and others Softbank.

• Perplexity debuted its new Computer system that it says features agents that can run for hours or even months on their own.

• Microsoft debuted new Copilot tasks that offer agents to help do work for users.

• Samsung announced a number of AI features for Android on the new Galaxy S26, including new Gemini capabilities powered by Google and integrations with Perplexity.

The Latest On Big Technology Podcast: Anthropic vs. The Pentagon, Bloodbath at Block, The Citrini Selloff

Ranjan Roy from Margins is back for our weekly discussion of the latest tech news. We cover: 1) The origins of Anthropic’s stare-down with the Pentagon 2) Claude’s use in the operation to capture Venezuela president Nicolas Maduro 3) Was Claude really being used for autonomous warfare or mass surveillance, and did the military seek it out? 4) Maybe this is just a culture clash 5) Anthropic’s marketing win 6) Should AI be used for autonomous warfare? 7) OpenAI raises $110 billion 8) Is that money real? 9) Block to cut nearly half its staff 10) Can AI be helpful in managing large companies? 11) Another science fiction story leads to a market panic.

You can listen on Apple Podcasts, Spotify, or your podcast app of choice

Join Big Technology’s Private Discord Server!

Where we’ll talk about this story, the latest in AI, the week’s podcast, and plenty more. You can sign up via the link below:

Read more

Scale AI, the AI data company that helps DeepMind, Meta, and Apple train their models, is now mostly working on reinforcement learning, a method that trains AI models to complete tasks, as opposed to traditional LLM training, which makes them more book-smart.

The shift is a stunning evolution in the direction of large language model development, indicating that traditional training has hit some limits while, at the same time, reinforcement learning — along with the agentic use cases it enables — is showing real promise.

“More than half of the training work we do involves reinforcement learning,” Chetan Rane, Scale’s head of product for agents and reinforcement learning environments, told me. “It was probably less than a quarter even six months ago.”

Traditional large language model training (also known as self-supervised learning) focuses on predicting patterns, while reinforcement learning gives models goals and lets them figure out how to achieve them. Current AI models are smart enough at the base level to attempt these tasks, and reinforcement learning is the method they’re using to figure out how to complete them proficiently. This task-based training is meant to transform AI from giving text or voice-based responses to doing things online.

For each task a model tackles, trainers must build highly-specialized, reinforcement learning ‘environments’ that mimic the steps the model will take when it’s let loose on the web. Teaching an AI model to navigate banking websites, for instance, involves building out user flows that account for almost every conceivable step and then setting the AIs loose within them. When the AIs figure out how to achieve their goals, the data is fed back into the models’ DNA — or weights — and it becomes more capable of accomplishing them.

There are some real limits on reinforcement learning though, including that the method tends to not generalize well, or handle scenarios it hasn’t seen before, even if its absolute level of intelligence goes up. “There’s proof that some amount of generalization happens,” Rane said. “But there are fundamental limitations.”

As AI training with reinforcement learning moves forward, expect the divergence between AI models to grow as the AI labs decide which types of activities to emphasize. One AI model company may make a model that is really good at accounting tasks, for instance, while another might be good at booking concert tickets. This is potentially one place where the AI race will sort itself out, as certain models become better equipped to handle certain tasks.

“We’re seeing a lot more divergence today than frankly, we did a year ago,” Rane said, “when everyone was just trying to build basic intelligence and keep up with one another. Now it’s clear that some labs are going more after the coding space, others are going more after the enterprise, others may be going after consumer.”

As these models roll out, don’t expect the fundamental economics of AI infrastructure to change though. Rane expects them to use much more compute, explaining that there’s a large number of calculations required for each attempt at a task. “RL requires much more compute than previous training paradigms,” he said.

Subscribe now

How Texas Instruments Builds Foundational Semiconductors and What They Enable (sponsor)

Jeff Morroni is the CTO for power at Texas Instruments. Morroni joins us live on location as Texas Instruments opens its brand-new semiconductor fab in Sherman, Texas. In this conversation, Morroni goes into depth about the different kinds of semiconductors that exist and what job they do, what the semiconductors built in Sherman are used for, and how innovation is at the core of TI’s technology development and manufacturing capabilities. Tune in for a fascinating, on-the-ground look at the reality of building the pivotal component in electronics today, and how TI is working to expand capacity now and for the future.

Watch Now

What Else I’m Reading, Etc.

OpenAI raised $110 billion from Nvidia, Amazon, and Softbank [CNBC]

Anthropic says no to the Pentagon [WSJ]

Jack Dorsey cuts nearly half of Block’s staff, cites AI, predicts others will follow [SFGate]

The Citrini note that crashed markets [Citrini Research]

Citadel’s takedown of the Citrini note [Citadel Securities]

AI’s Dr. Manhattan Syndrome [Person Familiar]

World Economic Forum CEO out after Epstein ties revealed [New York Times]

This Week On Big Technology Podcast: Can AI Become Conscious? — With Michael Pollan

Michael Pollan is the author of A World Appears: A Journey into Consciousness. Pollan joins Big Technology Podcast to discuss whether AI can ever become conscious and what that question reveals about the nature of mind. Tune in to hear a nuanced debate about whether consciousness is computable, where today’s LLMs fall short, and how researchers might actually test machine consciousness in the future. We also cover materialism vs. spirituality, the “hard problem” of consciousness, psychedelic experiences, and the emerging science of plant sentience. Hit play for a thoughtful, surprising conversation that brings the AI consciousness debate back down to earth while opening up some of its strangest possibilities.

You can listen on Apple Podcasts, Spotify, or your podcast app of choice

Thanks again for reading. Please share Big Technology if you like it!

And hit that Like Button to help reinforce Big Technology’s ability to keep sending this newsletters :)

My book Always Day One digs into the tech giants’ inner workings, focusing on automation and culture. I’d be thrilled if you’d give it a read. You can find it here.

Join Big Technology’s Private Discord Server!

Where we’ll talk about this story, the latest in AI, the week’s podcast, and plenty more. You can sign up via the link below:

Read more

Welcome back to Big Technology’s Agenda Setter email, in your inbox most Mondays with a preview of what’s to come in the week ahead.

Nvidia reports earnings this Wednesday and the stakes will be tremendous as usual. CEO and co-founder Jensen Huang is expected to give updates on inference demand, the pace of shipped compute, and more on the company’s vision for robotics.

In a rocky month for AI (and the software stocks it might destroy), Wall Street’s bar for Nvidia is especially high, with some expecting between 60% and 70% year-over-year revenue growth. Nvidia stock has barely moved since the start of last year’s fourth quarter, it’s up 1.7% vs. the S&P’s 3.3%, and it might take a legendary report to cause it to budge.

So even as the tech giants have committed to spending $600 billion on infrastructure in 2026, the company’s earnings results will be a must-watch as usual.

There’s also a new wrinkle for Nvidia this week. After months of reports that the company’s relationship with OpenAI has been fraying, we may indeed see it come through on plans to invest $30 billion in OpenAI, part of a $100 billion funding round that is reportedly nearing a close.

If the OpenAI round is not announced this week, Huang might still provide some clarity (and likely steadfast support) about how he feels about Sam Altman and Co. Those comments may well be the week’s big headline.

One thing worth your attention:

It’ll be worth watching how Huang talks about supply and demand. On its last earnings call, when an analyst asked him if there’s a “realistic path” for supply to catch up with demand over the next 12 to 18 months, he didn’t give a straightforward answer. More clarity this time would be nice.

We’re also less than a month away from Nvidia’s annual GTC conference, where Huang said the company will “surprise the world” with a new kind of chip. Any new information on that front could make waves.

Other events this week:

• Other companies scheduled to report earnings this week, with each likely to give updates on their AI roadmaps:

• February 24: Workday, GoDaddy, HP

• February 25: Salesforce, Snowflake

• February 26: Dell, Intuit, Coreweave

• The U.S. Senate Foreign Relations Committee on Tuesday will have a briefing to discuss “tech competition, critical minerals, and supply chain security.”

• The Senate Judiciary Committee will meet on Feb. 26 to discuss several proposed bipartisan bills related to protecting children online, including the COP Act, the ECCHO Act and the SAFE Act.

Subscribe now

The Intelligence Report

• Perplexity said it’s moving away from ads, claiming ads hurt trust in AI right as rival ChatGPT begins tests to introduce ads.

• Sam Altman and Dario Amodei refused to hold hands while appearing on stage at the India AI Impact Summit.

• ChatGPT’s early brands reportedly testing ads include Best Buy, Expedia, Qualcomm and Enterprise according to AdThena.

• Meta plans to spend $65 million this election cycle to fund AI-friendly state candidates.

• Bernie Sanders and Ro Khanna spoke at a Stanford town hall about billionaires and the future of AI.

• Chinese tech companies Alibaba and Moonshot released new AI models, which both feature text, video and image generation.

• Google and Apple announced new music focus features, with Gemini adding a way to create 30-second music tracks and Apple Music adding a way to make AI-generated playlists.

• ByteDance’s new AI-video generator Seedance 2.0 reportedly sent Hollywood into a panic, with the Motion Picture Association sending a cease-and-desist letter.

Your Mac: Calmer, Faster, and Tended From The Inside. (sponsor)

Your Mac works best when its inner space feels cared for. CleanMyMac helps you tidy clutter, prune outgrown files, and guard against malicious “bugs,” so everything runs smoother.

With CleanMyMac, you’ll see what’s taking up room, what’s slowing you down, and what’s safe to remove. Think of it as a calm, competent caretaker for your system, whether it’s for a quick sweep or a full tune-up. Use code Big20 and run your Mac better today:

Learn More

Under The Hood: Model Updates

Gemini 3.1 Pro:

• Google announced Gemini 3.1 Pro, which will power new capabilities for consumer apps like Gemini and NotebookLM along with various enterprise products and platforms. One of the things Google touted was its performance in benchmarks, including more than double the reasoning performance compared to the previous Gemini model.

• In a blog post, Google gave a few examples of how Gemini 3.1 Pro could power various tasks that benefit from advanced reasoning, including code-based animations and more advanced web design,

• Early enterprise clients testing Gemini 3.1 Pro include JetBrains, Databricks, Cartwheel and Hostinger Horizons.

• Some of Google’s updates seem to be positioning Gemini to further powers agentic workflows via data synthesis, long context, multi-step tools, and agentic coding.

Claude Sonnet 4.6

• Anthropic debuted its new Claude Sonnet 4.6 model, which it says is reaching capabilities similar to its larger Opus 4.6 model, but at Sonnet’s lower pricing. It’s also introducing Sonnet 4.6 with 1 million token context window in beta, which Anthropic says is big enough to hold entire “codebases, lengthy contracts or dozens of research papers in a single request.”

• Sonnet 4.6’s capabilities include improving agentic tools and upgrades in computer use that let the agent operate software like a human would. It also has better long context reasoning and better capabilities for business applications like financial analysis and document analysis. Although Opus 4.6 still beats Sonnet 4.6 in various benchmarks, the gap seems to be closing.

• Some of the platforms Sonnet 4.6 will update include Claude Code and Claude Cowork. When Anthropic asked users about their model preferences, it found users preferred Sonnet 4.6 over Opus 4.6, thanks in part to Sonnet’s better context awareness, fewer hallucinations and more consistent work with coding tasks.

Share

Mark Zuckerberg goes to court in Los Angeles

Meta CEO Mark Zuckerberg took the stand last week for a rare appearance in a civil trial that could alter how social media companies defend themselves in lawsuits alleging harm from platform design features.

The jury trial in Los Angeles, now entering its third week, is the first in a series of “bellwether” cases to test whether features like infinite scroll and algorithmic feeds can be treated as addictive product designs instead of neutral tools for organizing content.

Zuckerberg’s appearance came almost exactly two years after his 2024 appearance in Congress, where he infamously offered an impromptu apology to the families of kids targeted by social media abuse. Earlier in the trial, Instagram chief Adam Mosseri also testified and answered questions about whether engagement features can meaningfully contribute to addiction-like behavior.

The plaintiff in this case is a 20-year-old woman identified as K.G.M., who is one of roughly 1,600 plaintiffs in numerous cases alleging that Meta and other companies engineered engagement-driven features leading to compulsive use by minors and contributed to anxiety and depression. YouTube is also a defendant in the lawsuits, but TikTok and Snap — which were both initially included as defendants — settled before the trial began.

The case hinges on whether social media companies’ engagement-driven features can be treated as product design choices instead of editorial decisions. Historically, the latter has been protected by Section 230 of the Communications Decency Act, which has largely allowed platforms to avoid being held liable for content on their platforms.

The case could also raise issues about generative AI could pose new risks as features become powered by large language models and other types of AI-generated content. Even beyond the trial, there’s growing concern about potential psychological harm from chatbots like ChatGPT and Claude, including “AI psychosis” or other types of emotional manipulation and misinformation. Of course, Meta, Google and others have already come under scrutiny for their own LLM-powered chatbots.

Just like the early days of social media, the impact of chatbots is still unclear, but experts are already trying to understand potential causes and correlations. Last month, Harvard researchers published findings from their 2025 survey of 20,000 U.S. adults that found daily or frequent use of AI was associated with higher levels of symptoms related to anxiety and depression. They also found frequent AI use was more common in some groups including men, younger adults, people in cities, and even people with higher levels of education and higher incomes.

The latest on Big Technology Podcast

• Box CEO Aaron Levie joined the podcast to talk about OpenAI’s rumored $100B raise, Anthropic’s new model, and OpenAI’s acquisition of OpenClaw.

  “I think you’re in a period right now where you’re just in the infrastructure build out, teach the world about AI. It’s sort of worth subsidizing a lot of these use cases because it’s the fastest path to figuring out where the actual value is going to be.”

  -Aaron Levie

• Wrapping up Big Technology’s Davos podcast series, Google DeepMind COO Lila Ibrahim and SVP James Manyika outlined Google’s evolving approach to R&D amid the intensifying AI race. Alongside AI’s role in education, they discussed products like NotebookLM and Workspace, as well as longer-term bets in quantum computing, materials science and even training AI in space.

You can listen on Apple Podcasts, Spotify, or your podcast app of choice

Join Big Technology’s Private Discord Server!

Where we’ll talk about this story, the latest in AI, the week’s podcast, and plenty more. You can sign up via the link below:

Read more

Before AI, human writing at Amazon was sacrosanct. The company began each big meeting with six-page narrative describing the product or feature, typically written by the project lead, read in silence before anyone spoke. The writing’s purpose was to crystalize thinking and anticipate every scenario. Powerpoint, the enabler of logical leaps, be damned.

“The document should be written with such clarity that it’s like angels singing from on high,” Jeff Bezos once said. “I like a crisp document and a messy meeting.”

But now, Amazon is in its AI era, and its leadership is encouraging employees to let AI do the writing for them. The company’s internal marketing for Cedric, its ChatGPT-style tool, promises “six-page narratives in seconds.”

The implications for Amazon’s culture struck me as so profound that I reached out to over fifteen current employees, a mix of Bezos-era veterans and newcomers, to explore what the mandate says about their work lives, the company’s priorities, and what it means to be “Amazonian” today.

The worry within the company, I learned, is that Amazon is losing sight of writing’s centrality in its deliberative, thoughtful culture as it pursues powerful, new tools.

“Writing is thinking,” said one longtime company veteran. “That was the whole point of Amazon’s writing culture. I can’t tell you how many times I changed my mind when writing a narrative. And even when I didn’t, my arguments were more precise for having written them down. Now we have chatbots writing six-pagers to be summarized by other chatbots.”

Heavy Pressure to Use Inadequate Tools

Within Amazon, even those who expressed enthusiasm for generative AI described Cedric and other internal AI tools in terms like “comically inadequate.” For security reasons, Amazonians aren’t allowed to use external tools like ChatGPT.

“It has yet to return a single correct answer to me,” said one AWS employee, while another described it as “constantly hallucinating and then apologizing for being wrong.”

“I could use it to write docs, yeah,” said one self-described fan of human-centered AI. “But I don’t trust the output, so I’d rather just write it myself versus have to double-check every detail.”

Still, employees described heavy pressure to write documents via AI, whether or not the tools are up to par. “We get grilled in every team meeting about whether we’re using it,” said one junior Retail employee, while others receive weekly reminder emails from their VPs.

What they aren’t getting from leaders is an understanding of why machines should be doing their writing, beyond saving time or hand-wavy platitudes like “we should be using AI in everything we do.”

Employees are craving training and best practices for using it intelligently. In an open letter, CEO Andy Jassy encouraged Amazonians to attend company AI workshops and team brainstorming sessions. But when I asked employees about those, I got mostly head shakes or sardonic chuckles in response, suggesting that training is either spottily available or people don’t know it exists.

Amazon is a famously data-driven company, and the people I spoke with also expressed frustration that they aren’t hearing from leadership about how the impact of generative AI on the business is being measured qualitatively.

“There’s a communication vacuum,” said one Stores employee. “You can’t just say ‘Do it because I said so’ to hundreds of thousands of people with no real direction or training and expect to succeed.”

Amazon did not return a request for comment.

Resistance and Resignation

During these conversations, I began to mentally split employees into two general camps, the Resisters and the Resigned. The Resisters, many of them old-timers, see the AI push as both a symptom and an accelerator of declining standards. “I’m appalled at half-baked products being passed off as ‘polished’ work,” said one company veteran. “Trainings with incorrect information. Business reviews submitted with no narrative to put the data in context. [Documents] that the owner can’t even speak to. The list goes on and on.”

The Resigned tend to be newer and more junior, with less autonomy over how they work. “My manager is asking for longer documents now, because he says he can use AI to summarize them instead of reading the whole thing,” said a Retail employee who joined Amazon last year. “How is that an improvement over writing a quick one-pager that people actually read?”

Like other Resigneds, she’s complying for fear of falling behind if she doesn’t. “Expectations for volume of work keep rising,” said a product manager, “especially after these last few years of layoffs. If my teammates are using AI for their docs and I’m not, I’ll just end up in Focus [Amazon’s performance management tool] for failing to keep up.”

Other Resigneds say that expectations for quality of thought and communication simply aren’t as high as they used to be. One longtime product manager said it’s “screamingly obvious” when a document was AI-authored because it’s full of buzzwords and the kind of “excited marketing intern” language that used to earn a slap on the wrist from hype-allergic leaders. “A [written document] is basically just a box to check now. I’m not sure how much the quality really matters.”

Story continues below…

Promevo CEO Karthik Kripapuri on Google’s AI Comeback, Enterprise Agents, The Real Path to AI ROI (sponsor)

Karthik Kripapuri, CEO of Promevo, joins Big Technology to discuss how companies are finding real ROI today from cutting-edge AI models. Tune in to hear what’s changed on the ground with Gemini and Vertex AI, how companies find ROI by starting small with clear KPIs, and why data integrity is still the biggest blocker to agents delivering value. Hit play for a practical roadmap to what’s working now in enterprise AI and what it’ll take to scale it.

Watch Now

A “Grand Experiment”

As someone who found Amazon’s intellectual rigor rewarding and exhilarating in my twelve years at the company, all of the above made my head spin.

Worried that I might be sentimentalizing the past rather than understanding today’s Amazon on its own terms, I turned to two long-tenured executives for their perspective.

Read more

Elon Musk’s AI bot Grok is ascending, jumping from 1.6% to 15.2% in market share among daily U.S. users of chatbot apps from January 2025 to January 2026, according to mobile insights firm Apptopia.

The rise marks a furious entry into the chatbot app market for xAI, Musk’s AI company, which joined the race from a standing start in mid-2023. Grok now trails only ChatGPT and Gemini and it’s gained ground on both in the past year. Meanwhile, Grok has passed Copilot, Perplexity, DeepSeek, and Claude.

Grok’s growth story, however, isn’t completely clean. The app surged in downloads at the start of 2026, right when news broke that the bot would generate undressed images of people upon request, including minors.

The download spike took the app from around 500,000 daily downloads in late December to close to 1 million in early January, and lined up with Grok’s jump from 12.0% to 15.2% market share between December 2025 and January 2026.

Grok’s user base, perhaps unsurprisingly, is overwhelmingly male. Over the past six months, 82% of Grok’s weekly active users have been male, according to Apptopia. By comparison, males comprise 49.9% of ChatGPT, 45% of Gemini, and 77.8% of Claude users.

Grok has built features to specifically attract this audience, including an anime AI companion that, with some prodding, will engage users in sexual fantasies (NSFW). Musk has played into this, posting a Grok-generated video of a woman saying “I will always love you.”

The Washington Post recently reported that some xAI employees signed waivers to acknowledge their work would expose them to “sensitive, violent, sexual and/or other offensive or disturbing content.” They then reviewed sexual conversations users had with Grok, and worked to train the bot on those situations, the Post reported.

Adult companionship with AI bots seems to be viewed as an important engagement driver within AI companies. OpenAI is planning to roll Adult Mode in ChatGPT in the coming months, a feature built for users who would like to have spicy talks with their LLMs. Already, OpenAI’s pre-adult-mode GPT-4o model was so beloved among some users that the company caused a wave of grief by retiring it this week. “This will be looked back as a criminal offense in the future,” one bereft 4o user wrote on X.

OpenAI last month also fired a policy executive who spoke out against Adult Mode, though the company said not for that reason. The feature has real opposition within the company, the Wall Street Journal reported.

As the leading AI companies make their way toward the public markets, the pressure will be on to sustain growth and user engagement. That race intensified in recent weeks as SpaceX, which may go public as soon as early as June, acquired xAI, giving it a chance to beat OpenAI to the public markets.

Once these companies face investors quarterly, it will be harder to hold back the love bots if it means missing earnings expectations, or losing ground to the competition. And Grok’s strategy and growth may preview some of what’s to come.

A MESSAGE FROM OUR SPONSOR

From answers to action with agentic virtual agents

AI no longer stops at generating answers. Discover how autonomous AI can complete end-to-end workflows, not just one-off actions. With built-in, enterprise-grade autonomy, you can scale without sacrificing accuracy or trust. Take a closer look — read the blog.

What Else I’m Reading, Etc.

OpenAI fires product policy VP who spoke out against ‘Adult Mode’ [WSJ]

People are sad about losing GPT 4o. Imagine when ChatGPT gets more intimate! [TechCrunch]

The reality and privacy problems with Amazon Ring’s Search Party feature [GeekWire]

Salesforce CEO Marc Benioff made ICE jokes at a company kickoff event [Business Insider]

Norwegian Olympian wins medal, confesses cheating on his girlfriend in the TV interview [AP]

She doesn’t seem ready to forgive [Guardian]

This Week On Big Technology Podcast: Who’s Winning The AI Race? + Software’s Future — With Sridhar Ramaswamy

Sridhar Ramaswamy is the CEO of Snowflake. Ramaswamy joins Big Technology Podcast to break down the competitive dynamics in the AI race today, drawing from his experience working at Google and competing with it. We also cover the future of software, looking at whether AI will turn established software companies into “dumb backends.” In the second half, we discuss “shadow AI” driving enterprise adoption from the bottom up, the risk of becoming a feature in someone else’s platform, and why Chinese open-source models might actually be a net positive for the US. Hit play for a sharp, deeply informed conversation about where AI competition, enterprise software, and the future of work are heading.

You can listen on Apple Podcasts, Spotify, or your podcast app of choice

Thanks again for reading. Please share Big Technology if you like it!

And hit that Like Button… it’s good clean fun.

My book Always Day One digs into the tech giants’ inner workings, focusing on automation and culture. I’d be thrilled if you’d give it a read. You can find it here.

Join Big Technology’s Private Discord Server!

Where we’ll talk about this story, the latest in AI, the week’s podcast, and plenty more. You can sign up via the link below:

Read more

Welcome back to our Big Technology Agenda Setter newsletter, in your inboxes most Mondays from myself and journalist . This week we look at the Super Bowl ad fight, Nvidia’s forthcoming decision on OpenAI, and the market’s continued software conundrum.

The Super Bowl advertising fight between OpenAI and Anthropic might’ve been more interesting than the game itself.

Anthropic used its (very expensive) ad space to attack OpenAI’s plan to introduce ads in ChatGPT. The ads — one of which featured a therapist pitching a mature dating site to a young man having problems with his mother — certainly succeeded in annoying OpenAI’s executives. “The main thing that’s wrong with the ads is using deceptive ads to criticize deceptive ads,” OpenAI CEO Sam Altman said last week.

It’s unclear how much Anthropic’s ad got viewers to “keep thinking” though. An iSpot survey of 500 viewers placed the company’s ad in the bottom 3% of Super Bowl commercials over the past five years, according to Adweek.

Meanwhile, OpenAI used its airtime to introduce its Codex coding agent to the masses. And perhaps that wasn’t the best use of ad dollars either.

Overall, the AI ads were interesting to those in the AI bubble, but less so to those outside, underscoring just how much work OpenAI, Anthropic, and the gang must do to convince the broader public to use their products. Most people in the U.S. never use a chatbot app at all, according to data Big Technology published last week.

Anthropic’s ads, at least, likely had a target audience of enterprise AI buyers, and not everyone watching the game. It probably got those buyers to laugh. And that was a win in a Super Bowl that was mostly dull, both on the field and during the ad breaks.

Here’s a quick rundown of the other Super Bowl ads marketing AI — or made with AI:

• Google leaned into sentimentality with a Gemini ad about a mother and son using AI to imagine a new home after a move.

• Meta and Oakley emphasized “Athletic Intelligence” in an action-packed ad for its smart glasses while Microsoft also leaned into athletics to pitch Copilot.

• About a fourth of ads in the Super Bowl featured AI, with each taking a different approach in their marketing. (However, some online commenters said there were too many.)

• An AGI-themed ad from AI.com, a domain reportedly bought for $70 million by Crypto.com CEO Kris Marszalek, teased a new AI agent platform while offering few concrete details beyond a sign-up prompt.

• Other Super Bowl ads marketing AI came from Wix (for website building), Ramp (for automating expenses), Rippling (for workforce management), and Artlist (for AI content).

• Sprout Social’s social media analysis found “AI” was mentioned nearly 7,000 times in relation to Super Bowl ads between Jan. 27 and Feb. 9 on game day, with some chatter about various ads and others which ones were AI-made or not.

• The most mentioned AI-related Super Bowl ad was Anthropic with 6,000 mentions, followed by OpenAI (4,800), Meta and Oakley (2,400) and Gemini (500), according to Sprout Social.

• Coinbase, for whatever reason, decided to air a Backstreet Boys singalong.

Share

The Intelligence Report

• Big tech plans to spend $650 billion this year as AI race intensifies (Bloomberg)

• US companies accused of ‘AI washing’ in citing artificial intelligence for job losses (The Guardian)

• OpenAI launches Codex app to gain ground in AI coding race (Reuters)

• Voice AI Startup ElevenLabs Raises $500 Million (WSJ)

• “An AI bubble is not big tech’s only worry: Are Meta and Google ads really recession-proof?” (The Economist)

• Pinterest sacks engineers for tracking staff job cuts (BBC)

• New York lawmakers propose a three-year pause on new data centers (TechCrunch)

NVIDIA’s Definitive OpenAI answer

• News: Nvidia is reportedly nearing a deal to invest $20 billion in OpenAI as part of the startup’s plans to raise another $100 billion.

Read more