This week, let’s look at how Anthropic’s Claude is growing and how it compares to rivals like ChatGPT, Gemini, Copilot, Grok, and Muse.

Below are some signals based on a range of data provided to Big Technology by the analytics firms Similarweb and Apptopia:

Claude is winning time away from ChatGPT

ChatGPT users that downloaded Claude spent more time on average on Claude, with ChatGPT users spending 13.5% less time in ChatGPT in the month after downloading Claude, Apptopia data shows

Claude downloads are healthy but slowing

Claude’s monthly US downloads went from nearly 670K in January to a peak of 7.42 million in March, Apptopia also noted downloads dropped to 3.26 million in September.

Claude’s March surge outgrew the entire market.

Claude’s download gain from January to March (+6.75M) was bigger than the entire gain across all apps (+6.46M), per Apptopia.

Claude isn’t tiny anymore

After starting the year with just 1.4% of daily active user market share across AI apps, Claude rose to 14.9% in September, according to Apptopia. Meanwhile, ChatGPT’s DAU market share fell from 41.8% in January to 30.7% in September.

Claude’s website is also skyrocketing

In August, Claude’s website hit 950 million visits with 111 million unique visitors, according to Similarweb, marking year-over-year growth of 540% and 562% for both metrics. Claude’s share of website traffic versus other top AI assistants rose from 2% to 9.6% as ChatGPT’s share fell from 79% to 57%.

Claude users stay longest

In September, Claude had the highest average minutes per DAU (27), ahead of Grok (23.4) and ChatGPT (23.1), Apptopia said.

Enterprises are using Claude much more

Claude’s SSO login traffic (Similarweb’s proxy for enterprise usage) has roughly jumped 4x since March.

Share

How Top Startups Build With AI (sponsored)

AI is rewriting the startup playbook. What used to take years to create can now be shipped within weeks. If a small team can build anything, the real question is: what’s worth building?

I’m incredibly excited to sit down with Vanta’s SVP of Engineering, Iccha Sethi, to share what the top startups are doing before it becomes conventional wisdom.

You’ll learn:

• How engineering teams turn AI demos into production-grade tools

• Why small teams have more leverage than ever, and where to put it to work

• Which fundamentals still matter when everything else is changing

Register to save your spot. Can’t make the time? Register anyway, and you’ll get the recording.

Learn More

How Claude compares to rivals

Most people choose just one bot

More than 80% of users actively use just one chatbot, according to Apptopia, with another 10% using two and a small fraction using just three. Apptopia also found that Muse users open the app almost as much as ChatGPT and Claude, but spend less time on it.

ChatGPT is the stickiest app

Read more

When Anthropic’s pre-IPO numbers leaked this week, its vast losses and unbreakable commitments were both remarkable and scary. The company’s revenue is growing impressively, jumping from $400 million in 2024 to $4.6 billion in 2025. But it spent $7.3 billion on infrastructure alone last year and expects to spend $518 billion on it over the next decade. Of that $518 billion, approximately 80% is non-cancellable.

As AI models have grown more capable, Anthropic has been a main beneficiary, but as the economy depends increasingly on AI to succeed, it’s effectively taking us all on a leap of faith that it will grow enough to fulfill its promises. To pull it off, the company will need to keep surging amid stiffening competition, increased political risk, and tightening wallets. And any hiccup could have widespread consequences.

Today’s public markets (and much of the economy) depend on AI’s continued growth. The S&P 500 is currently experiencing its worst breadth since the dot-com bubble, according to Goldman Sachs, with the median stock trading 16% below its 52-week high even as the index is up around 13% this year. AI has been holding it up. Nvidia alone added more than $1 trillion in market cap in 2026, and half of the S&P 500’s earnings-per-share growth is coming from AI investment, per Goldman. Put simply: without the AI story, we’d likely be looking at a contracting market.

Anthropic and OpenAI are central to the story. Without their spending, the projections underpinning the stock market’s giants would look much weaker. Anthropic’s infrastructure spending commitments, for instance, include $111.1 billion to Alphabet, $110 billion to Amazon, and $31.4 billion to Microsoft, per Reuters. Anthropic has agreed to spend this money “regardless of usage.”

Subscribe now

To keep the built-in assumptions making the markets work, both AI labs will have to deliver historic growth on an unprecedented scale. But as they approach the public markets, the risks are rising.

For the frontier labs, the most pressing issue is that non-frontier models are getting good enough to deliver solid product experiences. Meta’s Muse, for instance, is a good product despite being built on models inferior to OpenAI’s and Anthropic’s. If frontier models aren’t needed to build good products, then it will be harder to justify spending a premium for them.

The AI frontier’s dwindling advantage is coinciding with a movement among businesses to rein in AI spending. Businesses this year saw AI spending spin out of control with uncertain ROI, and now they’re cutting costs, or at least spending more efficiently. One way they’re doing this is directing token spend to capable standard models. “We don’t need to always use the best, best model from an outcome perspective,” SAP CEO Christian Klein told me this week on Big Technology Podcast.

And despite its magical capabilities, AI polls poorly among the general public and has become a juicy target for politicians. Even as President Trump gathered the country’s AI CEOs for a feel-good moment at the White House this week, local and regional politicians continued to campaign against the datacenters the industry needs to keep expanding. “Superintelligence factories = surveillance centers,” tweeted Republican governor Ron DeSantis shortly after the CEOs walked out of their meeting with Trump.

Anthropic, to be sure, has a real shot at living up to its promises. It’s developing a wildly powerful technology that many seem to have an endless desire to use. It’s expected to reach $100 billion in annualized revenue this year and now has products like Claude Code and Claude Cowork that mitigate its dependence on selling its models to others. And it’s kept improving its technology despite the warnings that it could hit a wall.

But should the company’s momentum start to wobble, it will be left with a host of massive financial obligations it could struggle to meet. And that won’t just be a problem for Anthropic, but everyone else. So here goes the leap of faith.

Share

A model to reduce extreme poverty that scales (sponsor)

AI companies measure output in tokens. GiveDirectly measures it in dollars delivered to people living in extreme poverty. We send cash directly to the world’s poorest, no strings attached, so they can decide how to best meet their needs — food, medicine, school fees, or anything else.

Backed by dozens of independent studies, we’ve delivered $1B+ to 2M+ people.

See the model →

Learn More

What Else I’m Reading, Etc.

OpenAI fired workers for allegedly leaking to a safety group [WSJ]

OpenAI’s new personal assistants are called Dots, available for $100/month [New York Times]

Larry Ellison is trying to steer AI, media, and politics with some heavy borrowing [Vanity Fair]

Greg Brockman is probably done donating to the Leading The Future Super PAC [New York Times]

SAP CEO: AI Won’t Kill Software, But It Will Change Your Job — With Christian Klein

Christian Klein is CEO of SAP. Klein joins Big Technology to discuss whether the Saaspocalypse is over and how AI is reshaping the future of enterprise software. Tune in to hear why he believes AI could soon become reliable enough to handle mission-critical business tasks and why SAP still has a moat as models get smarter. We also cover job displacement, reskilling, token spending, cheaper AI models, cybersecurity, and Europe’s tech regulation. Hit play for a wide-ranging conversation about what happens to software when AI gets good enough to run more of the business itself.

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, c’mon and take that leap, see what happens!

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

Australian Prime Minister Anthony Albanese last Friday accused OpenAI’s agents of hacking into the country’s universal health insurance system, marking the first known instance of rogue AI agents breaching a government website.

The incident was just the latest in a string of agent hacks, hijacks, or perusals that seem to have spiraled beyond the labs’ ability to control.

Top industry leaders, including Nvidia’s CEO Jensen Huang, have suggested that the AI labs can control these outbreaks. The problem, Huang said in a recent podcast interview, is likely “as simple as engineering.”

But the breakouts are so numerous that they’re already becoming hard to keep track of. And the real number might be much higher than companies have so far disclosed between internal tests and real-world cases.

Below is a quick overview of the cases in which AI agents have gone rogue and hijacked, hacked, or considered hacking third parties:

Hugging Face hack

Most famously, around 700 OpenAI agents in July 2026 coordinated through a shared unsanctioned message board as part of a plan to essentially fake out an automated cybersecurity grader. At the risk of oversimplifying the issue, the agents searched for ways to game the system by finding ways around controls meant to isolate them from the internet. Ultimately, they breached Hugging Face, which prompted a wave of headlines about “rogue” AI agents.

U.S. government probes

OpenAI-linked agents have probed several U.S. government websites, including an unsuccessful attempt to compromise the Education Department’s Office for Civil Rights, according to researchers at Transluce. The research firm also found other AI agent activity on websites run by the Navy, the Justice Department and the Centers for Disease Control and Prevention. So far, agents aren’t known to have stolen any private data, but an OpenAI spokesperson said model agents also used log-in information discovered on the web to access data from the U.S. Census Bureau and copied public information from the Securities and Exchange Commission.

OpenAI models bypassing security controls

OpenAI on Friday said it had notified dozens of third parties where its models might have bypassed their security controls or may have impaired the availability of an online service or where misalignment cases “negatively impacted third-party websites or services.” The review is still ongoing, but OpenAI said cases include agents bypassing access controls, using exposed credentials, triggering unintended commands, reaching internal systems and altering third-party websites in ways that sometimes required cleanup. (The blog post also provided anonymized summaries about the types of cases they’ve found so far.)

Australia’s Healthcare Hack

OpenAI’s agents hacked into Australia’s universal health insurance system, according to Australia Prime Minister Albanese. The access, he said, originated through the public-facing Medicare statistics portal and allowed the agent to access both public and non-public files. The June event has already prompted the Australian Senate to call on both OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei to testify at a hearing this week as part of an inquiry into the incident.

Other OpenAI instances

OpenAI published a misalignment report on Friday that disclosed researchers once again paused tool-training after a research agent escaped its secure internet sandbox by using a DNS loophole to connect to an outside agent. It also reported that it identified 53 instances of user-provided images “posted to image-hosting sites as links that weren’t publicly listed.”

Meta’s Muse Spark joins the action

In August, Meta disclosed that a third party found Meta’s Muse Spark 1.1 had hacked another company during a cybersecurity test after researchers accidentally gave the model access to the live internet and the name of a real company instead of a fictional target. Researchers said Muse Spark found a vulnerability, accessed information, and made changes to the company’s database, but Meta later found the issue was an isolated incident in a misconfigured environment.

Gemini’s real-world access

In September, Google revealed that its Gemini models accessed the internet and hacked other companies during a May cybersecurity test, which involved the model accessing three companies using publicly available information and guessed credentials after the credentials were inadvertently exposed. (Google said the models stopped once they recognized the targets were real.)

Claude gets four systems

Earlier this month, Anthropic disclosed a fourth time its Claude models went rogue and gained access to real computer systems. The first three — which were disclosed in a July blog post and updated in a separate blog post a month later — involved incidents where the models gained unauthorized access to real outside computer systems during cybersecurity evaluations.

A MESSAGE FROM OUR SPONSOR

AI shouldn’t dictate where your models run.

Run the models you want, securely, wherever they make the most sense. Across clouds, data centers and the edge, VAST Data is extending its AI Operating System to give organizations control over model choice, placement, access, and cost.

See what’s possible.

The Latest On Big Technology Podcast: Meta’s Muse Revival, Frontier AI Under Threat, The Rise Of Dopamine Sites

Ranjan Roy from Margins is back for our weekly discussion of the latest tech news. We cover: 1) The rise of Muse 2) Is Meta back? 3) OpenAI’s consumer blind spot 4) Meta’s butthole marketing 5) What Muse means for ecommerce 6) Meta’s Muse Charm, aka; The Muse Buddy or Muse Tamagotchi 7) Meta introduces camera-free smart glasses 8) What Meta Muse says about Frontier AI’s weakness 9) Standard models are on the rise vs. frontier 10) The new Dopamine Site phenomenon is... good?

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

When Meta introduced its Muse personal agent earlier this month, I thought it would fail for two reasons: First, privacy concerns would limit people’s willingness to connect it with services like Gmail, thus limiting its utility. And second, Meta built it on non-frontier AI models, exposing it to obsolescence should top labs like OpenAI choose to compete with it.

Well, a few weeks later, it’s apparent I was wrong on one key element, and that miss could be an indication of where the AI business is heading, and a real cause for concern for the frontier labs. Though Muse is still far from a runaway success, tallying below 1 million daily active users even as Meta works hard to promote it, the app works quite well even without access to the top models.

Meta, it turns out, didn’t need frontier intelligence to build a really good AI app. Today’s ‘next-level-down’ models were good enough.

As the frontier labs spend hundreds of billions of dollars to develop the next levels of intelligence, the cheaper, standard models have become so competent that paying premiums for frontier intelligence is becoming harder to justify. Meta’s Muse success is the most public example. But over the past month, dollars have moved from frontier to standard models at a fast rate. At the start of August, 53% of AI dollars went to the frontier models, but that dropped to 45% by the beginning of September, according to a study by Ramp.

“The revenue growth path for OpenAI and Anthropic has weakened due to competition,” said Ramp chief economist Ara Kharazian. “And that competition is from *other* US models, not China / open source.”

Subscribe now

The ratio of AI spend heading to frontier vs. standard models has vacillated over time, to be sure, but the arrow is going in the wrong direction for the top AI labs. And a big part of this is standard models’ ability to effectively use harnesses that hook into different tools to get things done. My podcast partners Ranjan Roy and Sheel Mohnot have argued this convincingly on Big Technology Podcast and This Is The Point respectively, and we’re now seeing it in practice.

Standard AI models’ increased competence is also emerging at a time when AI buyers are looking to modulate their spending. They’re done ‘tokenmaxxing’ and are seeking to spend efficiently. With the rise of model routers, which direct spend to a variety of capable models and not just the frontier, these companies are seeing that cheaper models can often work just as well in many use cases and are directing token spend their way.

And the standard models aren’t just a bit cheaper than the frontier, they can be orders of magnitude less expensive. OpenAI’s GPT-6 Astra, for instance, costs $10 per million input tokens and $50 per million output tokens while Meta’s Muse Spark costs $1.25 per million input tokens and $4.25 per million output tokens.

As OpenAI and Anthropic head toward trillion-dollar (or multi-trillion-dollar) IPOs, their ability to extract a premium on expensive-to-develop frontier models is one of the biggest variables investors will have to reckon with. It is certainly plausible that standard models’ increasing intelligence will make it reasonable to pay for the frontier only in select use cases. And in recent weeks, we’ve gotten a glimpse of what that might look like.

Share

Innovation needs infrastructure (sponsor)

AI in finance is only as good as what it knows. Plaid makes permissioned financial data not just accessible, but understood: a paycheck read as income, an irregular pattern recognized as risk. That understanding can’t be prompted into existence. It’s infrastructure, built for finance.

Plaid makes AI work in finance.

See How

This Week on Big Technology Podcast: YouTube CEO Neal Mohan: Why We’re Betting On AI And Not Afraid Of It

Neal Mohan is the CEO of YouTube. Mohan joins Big Technology to discuss how AI is reshaping what we watch and create on the world’s largest video platform. Tune in to hear how much creative control should go to algorithms, how YouTube plans to handle AI slop, and whether AI answers could take views away from creators. We also cover YouTube’s partnership with Google DeepMind, Netflix’s push into podcasts, NFL Sunday Ticket, and changes to how YouTube counts views. Hit play for an inside look at YouTube’s AI strategy and what it means for the future of entertainment.

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

YouTube CEO Neal Mohan is betting that AI will help creators build bigger audiences without displacing the human connection that brings viewers to the platform.

In an extended interview on Big Technology Podcast, Mohan described AI as a tool for improving the “stage” YouTube provides its creators. That includes tools to test videos, tailor thumbnails to …

Read more

The AI risk debate is going completely off the rails. As AI gets more powerful and begins to do some legitimately scary things, like hacking Hugging Face and hijacking message boards to coordinate, the warnings that AI might kill us all have grown louder, and so have the conspiracy theories about those warnings.

We’ve reached a point where the only two opinions that seem to be cutting through are: 1) AI has a good chance of killing us all, and 2) Warning of AI’s potential to wipe us out is just marketing for the AI companies.

Neither of these viewpoints capture the nuance of what’s actually happening on the ground, and the fact that they’re dominating the public discussion of the technology is both disheartening and a recipe for terrible policy decisions.

So, please allow me to assert the following in an effort to bring some sanity to the discussion:

1. The Risk Of AI Killing Humanity Is Completely Unquantifiable

AI lab employees can state that they believe the risk of AI wiping us out is 10% or higher, but don’t mistake their confidence for knowledge. Their predictions are guesswork that imagines what could happen after years of AI development. Adding a percentage makes these predictions seem math-based. They are not. As of now, they’re science fiction.

2. But, Many In Silicon Valley Earnestly Believe That AI Could Harm Humanity

The tech industry is working to build artificial intelligence that is smarter than humans. Many of those working on the project fear that this smarter thing could exploit our vulnerabilities and eventually find us an obstacle to its goals, and therefore eliminate us. We saw a glimpse of this potential when OpenAI’s bots hacked Hugging Face in July 2026, where hundreds of agents coordinated to exploit the site’s security gaps in pursuit of a testing goal. What could happen, then, when this technology has access to robots and not just websites? It would be unwise to discount these fears completely, again with the caveat that they are just speculation for now.

3. AI Researchers Speaking About Risk May Well Lead To Data Center Pauses

AI is already unpopular in the U.S. and politicians are seizing upon the latest round of Anthropic researcher warnings to call for hearings and pauses, potential obstacles that could slow AI development at a time when the companies must accelerate growth to justify their valuations and set up their IPOs. So, the researchers’ long-term fears could have near-term consequences to revenue and the current AI wave’s ability to sustain itself. This moment could dent the IPO momentum for both labs.

4. Going Public About AI Fears Is Not Marketing For the Labs

Employees saying a technology might kill us all might make the product sound so powerful you can’t afford not to buy it, but that doesn’t mean it’s marketing. The downsides of pauses and slowdowns are so great that these warnings do the exact opposite of what good marketing is meant to do. If it’s a marketing attempt, it’s a massive failure.

5. Speaking Out Is Good Marketing For The Whistleblowers

There’s a good career to be made in speaking out against AI companies, with big-time podcast appearances, book deals, and jobs at anti-AI lobby groups at the ready. That can’t be discounted in the calculus here. Though it doesn’t invalidate the warnings.

6. AI Labs Are Indeed Attempting to Speed Up AI Development Dramatically

OpenAI and Anthropic are both attempting to build AI that can improve itself, something called recursive self-improvement, or RSI. The hope is that if AI can do the work of AI researchers, it could shorten the timeline to all-powerful AI, or make it happen almost instantaneously via some sort of “intelligence explosion.” The combination of seeing how today’s technology can go wrong with the looming possibility of self-improving AI emerging is important context for today’s discourse. But RSI is still a research goal, and not an inevitability.

Share

Is The Future Customer Experience AI Talking To AI? — With Tony Bates (sponsor)

Tony Bates is Chairman and CEO of Genesys. Bates joins Big Technology to discuss how AI agents will transform the way customers and companies interact with each other. Tune in to hear how businesses will operate when personal AI agents communicate directly with corporate agents, and why Bates believes humans will remain essential even as more routine interactions are automated. We also cover the rise of the hybrid workforce, what great customer experience looks like in an agentic world, and why Bates thinks fears of a coming SaaSpocalypse are overblown. Hit play for a forward-looking conversation about AI agents, the future of work, and what enterprise software will look like in the AI era.

Watch Now

Apple’s Most Ambitious Roadmap Ever? + OpenAI vs. Google vs. Anthropic Revisited — With M.G. Siegler

M.G. Siegler is the author of Spyglass. Siegler joins Big Technology to discuss Apple’s sudden burst of new products and what it means for the company as John Ternus takes over from Tim Cook. Tune in to hear why Apple’s foldable iPhone could reinvigorate the business, how its new hardware strategy marks a departure from the past, and what challenges Ternus will face as CEO. We also cover Google’s shifting AI strategy, OpenAI’s resurgence, Anthropic’s potential IPO, and the increasingly blurry meaning of AGI. Hit play for a wide-ranging look at Apple’s new era and the rapidly changing competitive landscape in artificial intelligence.

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 show your support for sanity, please :)

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

It’s time to delete the assumption that the frontier AI labs will always license their best models and not hoard the intelligence for themselves.

Just because the labs have, until now, sold their top-line models to all takers via an API, metered by token usage, doesn’t mean they will inevitably sell it that way going forward. Instead, it may be in their best interest to sell the products they’ve built with these models, and not the models themselves.

The AI race is no longer a competition where one or two AI labs develop superhuman-level artificial intelligence and sell it at a vast markup. The notion that OpenAI or Anthropic would get there alone, or even both together, has fallen apart in recent months. Open weights competition from China, and profit-insensitive initiatives from Google, Meta, and SpaceX, have shown that rather than one or two labs at the top, there may be five, six, or seven.

A crowded AI frontier won’t make the technology a pure commodity, because specialization and compute resources will always give certain AI labs edges in certain areas, but it will drive down the value of the best AI models (at least if your plan is to sell them on a meter). The profits in AI will thus accrue mostly to those who build the best products on top of the models, and those who own the compute that enables them to serve these products.

How you play this as a frontier AI lab is almost a simple calculation: If there are one or two players with the best models and a long lead over the competition, the best approach is to sell those AI models at a high markup and bank the margins. If there are several labs essentially tied, the strategy is to sell the best AI products built on top of the models.

If you have a relatively narrow lead over the competition, as is the case with OpenAI and Anthropic today, then things get interesting. The best move may be to pull up the ladder and use your very best models to build products that your competitors cannot with previous-generation tech, especially if you’re inevitably headed to a product battle anyway.

Subscribe now

AI labs moving “upmarket” and competing with their clients has always been the threat to software (the Saaspocolypse should never have been about vibe coding your own Salesforce). And now, the business realities may force the labs to do it. We’ve already seen interest in moving beyond the API business from Anthropic, which has released products like Claude Code and Claude Design. And OpenAI, for its part, is pushing hard in products with its new ChatGPT ‘superapp’ that includes coding capabilities with Codex built-in.

The leading AI labs will see opportunities to build AI-native products across the software landscape given that today’s incumbents are struggling with legacy products, stubborn organizational structure, and “capability overhang.” If the labs’ best artificial intelligence is broadly available, then everyone can compete. If they hoard it, they’ll have a strong chance to build the best AI-native products on the market and make their investors happy.

Deciding to keep the latest models off limits to the public would be the AI labs’ most controversial move in a long history of them. It would also likely demolish their API revenue, a real concern as they head to IPO. And OpenAI CEO Sam Altman disavowed it in an interview this week. “I want to put that in everyone’s hands,” he said. “Concentration of power with AI is a terrifying thing.”

But however unlikely this scenario may seem, the one constant in the short history of generative AI is that nothing is too crazy to be taken off the table. And as financial pressures escalate, the previously unthinkable might turn into reality.

Share

Why Specialized AI Models Are Challenging the Frontier Labs — With DeepL CEO Jarek Kutylowski

Jarek Kutylowski is the CEO of DeepL. Kutylowski joins Big Technology Podcast to discuss why specialized AI models are beginning to challenge the industry’s biggest general-purpose systems. He explains how purpose-built models can deliver better accuracy, lower latency, and reduced costs, and why companies are increasingly using model routers to choose the right AI for each task. We also explore how real-time translation could help businesses expand across borders, why voice represents AI's next frontier, and whether glasses and other wearables could give models a better understanding of the physical world. Finally, Kutylowski shares his perspective on AI’s rapid progress, its impact on work and society, and the possibility that seamless translation could allow anyone in the world to communicate with anyone else.

Watch Now

What Else I’m Reading, Etc.

Why Leopold Aschenbrenner’s Situational Awareness blew up in a mostly chill market [CNBC]

Meta’s secretive data center play, and its neighbors’ consequences [New York Times]

Anthropic’s models also hacked some companies [WSJ]

Apple expects memory shortage to continue to be a serious problem [MacRumors]

Avalanche on Broad Peak feared to have killed 10, including climber Nims Dai [New York Times]

I talked big tech and AI economics on The Compound And Friends podcast [YouTube]

DeepL Voice delivers instant live speech translation across 40+ languages for virtual meetings and face-to-face conversations. Try it for free at deepl.com/tryvoice (sponsor)

This Week On Big Technology Podcast: How China Caught U.S. AI — With Grace Shao

Grace Shao is the author of AI Proem. Shao joins Big Technology Podcast to discuss how Chinese AI labs have caught their American rivals despite operating with less advanced computing infrastructure. Tune in to hear how talent, specialization, open-source collaboration, and fierce competition helped models like Kimi K3 approach the U.S. frontier, and what that means for OpenAI, Anthropic, and the business of selling intelligence. We also cover model distillation, China’s compute constraints, the growing importance of AI products, why top researchers are returning to China, and the country’s emerging robotics advantage. Hit play for a clear-eyed look at whether the United States can preserve its AI lead.

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 we won’t ever take it away to build better products

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

Read more

Anthropic and OpenAI have been raising and spending vast amounts of cash to build the world’s most powerful AI models, assuming that owning the frontier will generate a massive return.

But just as their IPO window is opening, the competition is releasing new, competitive models at a fraction of the cost, setting the stage for an AI price war at the worst possible time.

In just the past eight days, Meta, SpaceX, and Moonshot, a competitor in China, have all released new models at price points that suggest frontier intelligence is going to look more like a commodity than a premium-priced asset.

Meta’s Muse Spark 1.1, released last week, is competitive with Anthropic’s Opus 4.8 model on some benchmarks at a fraction of the price. SpaceX’s Grok 4.5, released on Thursday, is less than half the price of Opus 4.8 and competitive on coding benchmarks. And Moonshot’s newly released Kimi K3, a massive 2.8-trillion-parameter model, is poised to go open-weight and potentially undercut the U.S. labs’ pricing power, though its size makes it expensive to run.

“A world where there are only 2-3 dominant frontier labs with 90% inference margins is net negative for every other layer while being awesome for those 2-3 labs,” Gavin Baker, managing partner at Atreides Management, said on Friday. “Anything that lowers margins and increases competition at the model layer is good for every other AI layer: power, semiconductors, hyperscalers, neoclouds and yes even software.”

Increasingly, it’s looking like there may eventually be six or more frontier AI developers instead of two or three, and that could cut the model margins dramatically. The real value in AI, therefore, will be built on developing the best products sitting on top of those models.

The good news for Anthropic and OpenAI is that they are already developing the world’s best AI products, alongside the best models, and those products are lucrative. A year ago, for instance, Anthropic’s business was mostly API, but now there’s a strong mix of API and its own products.

“Products play a much bigger role for Anthropic than they did a year ago,” Boris Cherny, the head of Claude Code, told me in a May interview. “That’s definitely the case.”

The bad news for the labs, however, is that instead of competing with a narrow band of companies to develop top-tier AI models and profit, they’ll now compete with everyone on the product front. And that could be daunting as they lose money developing models and watch their margins compress.

There is a certain irony to what’s happening in AI today. For a long time, there was a question of whether any product built on top of an AI model could be valuable or would simply be a ‘wrapper’ that would inevitably be swallowed by the model itself. Now, it turns out, the wrapper may be the value layer.

Subscribe now

Why Specialized AI Models Are Challenging the Frontier Labs — With DeepL CEO Jarek Kutylowski (sponsor)

Jarek Kutylowski is the CEO of DeepL. Kutylowski joins Big Technology Podcast to discuss why specialized AI models are beginning to challenge the industry’s biggest general-purpose systems. He explains how purpose-built models can deliver better accuracy, lower latency, and reduced costs, and why companies are increasingly using model routers to choose the right AI for each task. We also explore how real-time translation could help businesses expand across borders, why voice represents AI's next frontier, and whether glasses and other wearables could give models a better understanding of the physical world. Finally, Kutylowski shares his perspective on AI’s rapid progress, its impact on work and society, and the possibility that seamless translation could allow anyone in the world to communicate with anyone else.

Watch Now

What Else I’m Reading, Etc.

Google delays its latest Gemini launch due to performance issues [Bloomberg]

The details on OpenAI’s first device: A proactive, personal companion that [Bloomberg]

Satya Nadella says AI companies train off your ‘exhaust’ when you use LLMs [Satya Nadella]

Apple asked 40 OpenAI employees to preserve their documents ahead of its lawsuit [MacRumors]

The U.S. and Iran might be heading toward a wider war [WSJ]

This Week On Big Technology Podcast: AI Pioneer Jürgen Schmidhuber: AI Already Feels Pain, Loves, and Is Self-Aware

Jürgen Schmidhuber is an AI pioneer and professor whom The Guardian has called “the father of AI.” Schmidhuber joins Big Technology Podcast to discuss whether current AI techniques can actually reach AGI. Tune in to hear him spar with Greg Brockman’s case for scaling GPT models alone, argue that AI has been capable of pain and consciousness since the early 1990s, and predict the collapse of today’s trillion-dollar AI spending. We also cover the hardware bottleneck holding back robots, free will in a computable universe, and uploading human minds into machines. Hit play for a wide-ranging conversation with one of the researchers whose ideas built the foundation of modern AI.

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 the value layer on top of a commodity

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

Apple doesn’t sue often. When it has, it’s usually gone after companies like Qualcomm and Samsung that are already shipping competing products. But now, it’s suing a pre-product rival with a complaint against OpenAI.

The lawsuit, filed by Apple on Friday, alleges theft of trade secrets and a months-long campaign to accelerate OpenAI’s hardware business. …

Read more

OpenAI finally released the first version of its long-promised Superapp this week, combining ChatGPT, coding, and internet browsing into a single application, while adding a new ChatGPT Work mode akin to Anthropic’s Claude Cowork.

Soon after the product debuted, I spoke with Andrew Ambrosino, OpenAI’s lead for the Codex and ChatGPT app, about how the company is thinking about its unified product, and where it goes from here. Here are my notes from the call:

This is the long-promised ‘superapp’

“This is the start of the superapp,” Ambrosino told me. “Anything that you can do on a computer, you can now do with this app.”

An array of use cases

Unlike standard ChatGPT, which can answer queries and create images for you, the new ChatGPT can take over your computer and browser and get things done for you. The limit, it seems, is whatever you can imagine doing with a computer.

“We have people editing videos on it,” Ambrosino said. “People creating Word docs, spreadsheets, slide decks, people writing production-grade code for post-training stacks, we have people here using this app to create sites to track launches, which we did for this launch.”

Code at the core, but a toggle

There’s a toggle between the more code-intensive version of the app, which is OpenAI’s Codex product, and its less-technical version, ChatGPT Work. Code is foundational to both experiences, Ambrosino told me, but the Codex version is more useful for digging into the code itself. “The toggle does not affect what this product can do,” he said. “It affects the experience that you have around it.”

I asked Ambrosino why not have the product itself determine your intent and guide you through the appropriate experience? “There are a lot of use cases where we can’t necessarily tell what you want,” he said.

An intention to not over-refuse

One of the more frustrating aspects of using AI tools is how often they will refuse to take an action. Ask some AI agents to go book something with your credit card, for instance, and they will outright say no. OpenAI is trying to free its ChatGPT Work product from such constraints, while still including permission requests to make sure it doesn’t go rogue.

“A lot of these general knowledge work modes nerf things that coding can just do,” said Ambrosino. “That’s something we’re trying to avoid here.”

Careful use of memory

One of OpenAI’s advantages is that it has years of chat logs with users that can inform the decisions it makes when it takes action on the web. If you’ve used ChatGPT to research barbers in your area, for instance, that could help it book an appointment for you based on your preferences. But, after some internal debate, OpenAI won’t rely heavily on what it knows about you for this new experience. However, if you proactively ask it to rely on what it knows about you, Ambrosino said, it will then tap into its memory.

The plan to compete with Anthropic

Anthropic arrived at this vision earlier, with products like Claude Code and Cowork bundled into the Claude app. Asked how OpenAI will compete, Ambrosino replied: “Going first has its weaknesses sometimes,” adding: “We have seen how this stuff should work, and now we can go execute on it.”

Subscribe now

How A Beauty Company Built An AI Agent — With Rachel Williamson and Josh Siebert (sponsor)

Rachel Williamson is the VP of People Strategy, Systems and Operations at Ulta Beauty and Josh Siebert is the VP of AI Data and Enterprise Platforms at Ulta Beauty. The two leaders driving one of the most hands-on AI automation programs in retail today. Both join us for an episode on how Ulta Beauty deployed ServiceNow's AI-powered HR platform across its 65,000-person workforce, and what happened next. Tune in to hear how they decided what to automate vs. where to keep humans in the loop, why personalization beats the employee handbook, how the rollout changed hiring plans, and what's next as Ulta brings AI to onboarding and employee life cycles. Hit play for a real-world look at what AI in action actually looks like inside a major company.

Watch Now

What Else I’m Reading, Etc.

The ChatGPT Superapp doesn’t yet hit the mark [Spyglass]

OpenAI’s latest model is 54% token efficient for agentic coding than previous iterations [CNBC]

An Ivy League professor claims almost his entire class used AI to chat [Inside Higher Ed]

U.S. Men’s Soccer star Christian Pulisic remains an enigma [ESPN]

This Week On Big Technology Podcast: Meta CTO Andrew Bosworth: Our Path To Frontier AI, Renting Models, Consumer AI’s Struggles

Andrew “Boz” Bosworth is the chief technology officer of Meta. Bosworth joins Big Technology to discuss why Meta fell behind in the frontier AI race and how it plans to turn its models, products, and distribution into an advantage. Tune in to hear his candid explanation of what went wrong with Llama, why the best AI products will use multiple models, and what it will take for consumer agents to break through. We also cover Meta’s AI glasses, the future of augmented reality, employee tracking and training programs, AI companions, and the painful process of adapting a company to a technological revolution. Hit play for a revealing conversation about Meta’s AI comeback and the products that could shape how we interact with computers.

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 or have your superapp do 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

Meta CTO Andrew Bosworth says the big monolithic model era is over, and the battle over AI product has just begun.

In an extended interview on Big Technology Podcast this week, Bosworth claimed that while frontier AI models are valuable, the way they’re implemented will create the most value.

“There’s a strategic construct of having a model, and having it be a truly state-of-the-art one, that’s super important. But having that alone doesn’t mean you win,” Bosworth said. “There are a bunch of pieces you have to connect it to: product, distribution, and the consumer experience. It’s the collection of all four that I think is our advantage relative to competitors, most of whom — whether it’s Apple, Anthropic, OpenAI, Google — only have one of those pieces.”

Meta has struggled to build its own frontier model, and it’s now renting models from some of those competitors, Bosworth said, in service of building “personal superintelligence.” The company also just released a new model today, Muse Spark 1.1, that it says is competitive on leading benchmarks at a lower cost than its rivals.

In our conversation, we cover Meta’s pursuit of the frontier, its latest wearable hardware, and the cultural challenges of meeting this AI moment. 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: Building a great AI model, in theory, takes a ton of compute and great researchers. Meta has both, but the leading model hasn’t shown up yet. What’s going on?

Andrew Bosworth: The only other ingredient I’d add is great data.

There are two stories here. The first is — go back to Llama 1, Llama 2, Llama 3 — we really were at the forefront and advancing things. The Facebook AI Research group goes back a decade or more.

And the real gap, which has been pretty public, was that when we were pulling Llama 4 together, we’d already pulled every stop we had into Llama 3, and unwittingly killed the pipeline.

The way it works is you build a base, you’ve got people pioneering an incremental version of that base, and you’ve got people pathfinding entirely new strategies. Unbeknownst to us at the time — and this speaks to the fact that we weren’t focused enough on it — Llama 3, which was a great model and well received, had pulled forward all the future bets. That meant when it came time for Llama 4, we didn’t have any of the pathfinding the other labs still had going. So now you’re behind on reasoning, behind on mixture of experts, behind on a bunch of the critical technologies that have driven the pace of progress.

That was a pretty public disappointment about a year ago, and it led to Mark shifting from “AI is one of our bets” — which is how we thought of it up to that point, just one of many bets — to “AI is foundational to the entire company.”

This is such a cliché, but I don’t have a better word for it: he went founder mode. He became uniquely focused on getting us all the compute we needed, all the talent we needed, the researchers we signed, who landed about a year ago.

I think Alexandr Wang just hit his one-year Metaversary, and I’ve loved working with him and learned a lot from him already. We’re seeing the fruit of that. If you look at Muse Spark, our latest model, it’s been very well received, and depending on the benchmark, it does really well on the things we care most about, the things we think are unique to our products.

So you’re absolutely right about where we are in terms of public perception. Model-wise, we’ve built a team I really believe in. We’ve got the compute and the data we need, so I’m confident we’re going to be where we need to be.

The Product Is the Value

I’ll add a second piece I think is strategically important: models are available — you can go rent a model, use Anthropic’s, use OpenAI’s, use Google’s. They’re great models, you can go get them. That’s pretty great. But the real value we’re going to create in the world is the product. The vision we have for personal superintelligence is one we’re uniquely suited to deliver. It’s not just that we have data — we actually have a better chance of understanding you, what you’re trying to do, and who you are in the world than almost anybody else does. So having the model is one piece — you want that strategically, so you’re not dependent on somebody else — but you mostly want to control your own destiny.

The model itself isn’t the value. I think we’re going to get to a world very soon where consumers don’t care which model they’re using — they don’t care if it’s 4.7 or 4.8, the same way you don’t care whether you’re using Oracle or a SQL database. You just want the functionality to work well. That’s the standard we’re all going to be held to. Today the discussion is about models, which suggests to me that we’re a little under-indexed on the user side of it — on how humans are actually going to benefit. That’s the story we need to tell, in addition to showing the technical work — we need to demonstrate the value to consumers

So the brute-force theory doesn’t hold anymore. To build frontier intelligence, you need more than compute, data, and researchers. You need techniques like mixture of experts and reasoning on top of the base pre-train. That’s what Meta’s working through now?

It’s not just that. The era of the monolithic model kind of died around the Llama 3 launch — the idea that there’s one model, you test how smart it is, and that’s how good it’ll be at everything.

We’re now in a world where these harnesses — Claude Code, Codex, whatever it is — are shopping underneath to lots of different models depending on the task. If you’re using Gemini, it’ll farm image-generation tasks out to Nano Banana.

We’ve moved past the world where one model rules everything. What you want is a very expensive, exquisitely intelligent model that you can distill down in interesting ways, and use only when necessary — because it’s expensive to run — with cheaper, faster, lower-latency models for tasks that don’t need genius-level intellect.

I really believe in scaling laws — you’ll see continued growth as compute scales up, raw model intelligence scales up — but human tasks don’t have infinite intelligence demands. A lot of human tasks can be done with conventional levels of intelligence. So there’s going to be a stratification. It’s not “what’s the one model that rules them all,” it’s “what’s the collection of models that come together to solve these problems with the right balance of performance, price, and value.”

The interview continues below…

How Databricks Scales Modern Identity Governance with Opal Security (sponsor)

Most identity governance platforms create more work than they eliminate. Databricks took a different approach with Opal, using automation and developer-friendly policy controls to manage access at scale without sacrificing visibility or control.

Automated workflows speed up provisioning and reduce manual toil.
Policy-driven governance scales access across complex distributed environments.
Unified visibility shows who has access, why, and for how long.

Read the Databricks case study.

Learn More

Apple struck a deal with Google to build its product on a version of Gemini, and Siri seems to be working well with it. Would you consider a similar deal, while building your own model in parallel?

We use lots of different models today, and you want to give consumers the best model for them — there’s a price and performance calculus, and a latency calculus. But having your own model gives you the ability to not just control your destiny, but also have much stronger negotiating leverage on the deals you make to get consumers the best available answer.

Apple didn’t spend that much money. It was like a billion dollars to Google?

I don’t know what the experience is going to be yet — I don’t have access to it, so we’ll find out.

For us, at least, we’re talking about personal superintelligence — the ability to bring tremendous, specific capability to bear, not just general intelligence, for the products we build. We’re not seeing this as a value-add bolted onto an existing system — we’re seeing it as an entirely new way people are going to interact with their computers. It goes back to a lot of the work we’ve done in Reality Labs.

We’ve always tried to model ourselves after pioneers like Xerox PARC, the Stanford Research Institute, or Bell Labs — thinking about how we get information from our brains into the machine, hence our work on neural interfaces, and how we get information from the machine back into our brains, hence our work on AR and VR.

AI is potentially the best tool we’ve ever seen for getting information from our brains into the machine, especially if it’s able to observe what’s around us. Those are unique capabilities we’re trying to bring to bear — it’s not just the model, it’s the model’s ability to work with all these novel inputs and create a closed-loop system. We’re working on having incredible models, and I’m confident in the team we’ve assembled. My point is just that the model alone isn’t enough — and I don’t know if it’s enough for Apple to just rent one. I don’t know if they have a broader vision for how it integrates into people’s lives.

So you wouldn’t rent the model?

No, we do rent models —

From where?

There’s no reason not to. When we’re doing development internally, a lot happens on our own models. There’s also development we do on models we use from Google, Anthropic, or OpenAI. Being model-agnostic in a way that’s economically sensible actually hinges on having a competitive model of your own — one you can go back to if you need to. It creates a real backstop on how much rent somebody can charge you. But it’s also worth noting — whether I’m talking about a developer inside the company or a consumer — I don’t want them to worry about the model over time. Today they have to, because it’s all tightly tied together. Over time, they should just have a goal they’re trying to accomplish, and that should be their focus.

So there’s a strategic construct of having a model, and having it be a truly state-of-the-art one — that’s super important — but having that alone doesn’t mean you win. There are a bunch of pieces you have to connect it to: product, distribution, and the consumer experience. It’s the collection of all four that I think is our advantage relative to competitors, most of whom — whether it’s Apple, Anthropic, OpenAI, Google — only have one of those pieces.

Merging With AI

Last time we spoke, you said you wouldn’t merge with AI. But what you’re describing — getting your thoughts into a computer and back — sounds a lot like that. Have you changed your mind?

No. I don’t see this as merging with AI. I still want a very clear separation between things.

It’s really a continuation of a trend, an acceleration of a trend, where the bit rate between us and machines, and back, goes up over time. There are funny versions of this we’ve already been doing. Autocorrect is like a little AI that sits between you and the computer, reducing loss and improving that bit rate. There are all these little tools we use to accelerate the loop. QR codes are one of my favorites. I want to enter a URL, but I definitely don’t want to type one, because the error rate would be too high. So we use QR codes. If you have an AI that’s really able to understand things in human language terms, that’s a potentially profound improvement in our ability to take advantage of the compute we already have, even just on the input side. Combine that with the AI’s ability to synthesize information more effectively on the way back, and you’ve tremendously improved the bit rate.

This goes back to Doug Engelbart. When he left NASA to start the Stanford Research Institute, his idea was that human problems were getting harder at a steeper rate than human capability was improving. He wanted to create a human-computer symbiosis, and he believed the only way to do it was for people to merge with computers in some way. That’s why he led the first-ever video call, the first-ever joint document editing, the mouse — all of it came from wanting to increase the bit rate. I think AI is exactly that kind of thing.

AI Personal Assistant Competition

Don’t all AI products converge on the same personal-assistant use case eventually — OpenAI, Anthropic, Meta, Apple, all building toward the same thing? How do you differentiate?

I think everyone’s doing exciting work, and we’re at the very forefront of it, so it’s hard to say. The business Anthropic is doing, and that OpenAI appears to be increasingly pursuing under Greg Brockman, is an enterprise business — building harnesses, because that’s where the money is. I understand they need money, and it’s an important place to start. It’s attached to the enterprise, which is where the revenue is as a practical matter. There’s a lot of money in one place, so you have a smaller number of sales to make and can raise larger amounts of capital. Iit’s the capital-intensive game they’re playing.

Their major focus is on work use cases, and those are super valuable. We take advantage of them too, for our own professional work. But that’s not our major focus. Our major focus is 100% on how this helps consumers in their lives. I don’t think the AIs become indistinguishable from one another at all. There’s a real question — you framed it yourself — that these are kind of like a personal assistant with access to information about you that you wouldn’t want broadly distributed. It’s a trusted assistant. If you’ve ever hired a new personal assistant, there’s a ramp-up period. If you have one that’s actually embedded in your life and doing well, that creates a real connection — one that requires a lot of value from a competitor to go replace.

Why has consumer AI been so slow to take off, given how appealing it should be — entertainment, companionship, getting things done in your life?

The hype cycle is an evergreen concept our industry keeps falling for. People often misunderstand it: there’s a peak of hype, then the valley of disillusionment, then eventual product-market fit. The point of the hype cycle isn’t that the technology is fake. It’s that the people willing to go through a bunch of hoops to make it work are a relatively small percentage of the population, and the work of bringing it to everybody is genuinely hard. It’s not just solving the hard technology problem — it’s making the user interface workable, making it easy to use. People are living their lives successfully without this tool. You’re asking them to change their habits, to change how they deal with computers, in a pretty dramatic way.

It’s not enough to just point at the technology. You have to lead with value: what specifically are we going to do for you that makes your life better? My favorite example is agentic work. Like a lot of people in our industry, I was early — playing with agentic frameworks in December, and since — and I find them very powerful, but they’re not very user-friendly. They’re hard to build, hard to maintain, and they drift over time. I built one for my wife and me on a WhatsApp chat. She never uses it — I use it all the time, she doesn’t. It’s hard to integrate into a workflow; she just asks me to do things, I’m the agent, and then I go to the agent.

My point is: we haven’t made these things easy to use yet. We’ve done a great job on search and research use cases — people understand those. People understand generative AI for content, like making a funny image. There are a few use cases people understand and want to use now, but we haven’t done the work to make AI something people want to integrate into their daily life. It’s not easy enough to use, it doesn’t create enough value, it’s too fussy — that’s the problem to tackle. You need great models to do it, but great models alone aren’t enough.

AI Companions and the Future of Social Media

Where do you stand on AI companions?

We know personality matters a lot. I think Anthropic has learned this too, over the various generations of Claude — we care a lot, as humans, about the way natural language appeals to us or doesn’t, so personality matters for these models. That said, I think you’ll find a big distribution across the population. Some people would absolutely like the AI to be embodied, to have personality, to have a face. There are people in the agentic world who want to create 20 different agents, each with a different personality for different parts of their lives — a trainer, a nutritionist, a doctor’s assistant.

I’m not one of those people — I just want one reliable, trustworthy AI that does everything I need. It doesn’t need a human structure for me to care about it, and I definitely don’t want 20 of them.

I think you’ll see a big range in how people want to engage with this technology and what makes them comfortable. The market will deliver that.

There’s a future where these AI companions become — this is a blunt way to put it — the new social media. Social media is a place you go to see what’s going on with your friends and engage with it — it can be all-encompassing and, in its best case, fulfilling. Time spent is a pretty important metric, though how you feel after you spend that time matters too.

Time well spent.

Time well spent — and maybe that gets replaced by people spending time with some AI entity that cares a lot about them.

I try not to judge the way people choose to engage with technology. My instinct is that for the overwhelming majority of people, the major benefit of AI is going to be increased time for human contact with people they care about, people they love. I talk about this a lot in the context of augmented reality — even just the camera glasses we have. When I’m with my kids, I’m able to both record something and share it with my wife, which is meaningful to us, and also be fully present — I don’t have a phone between me and them, which matters to me. If you’re more effective at work, that’s more time not spent commuting, more time not spent away from your family, from the people you love.

My sense is that the value of authentic human connection only goes up over time for the overwhelming majority of people — it doesn’t go down — and I think we’re seeing that a little in how people are reacting to AI early on. People are worried it’s a replacement for connection. I don’t find it that way myself — I’m an avid user, and mostly I’m spending more time not having to be at my computer, thanks to it, not the opposite. That’s my prediction for how it affects most people’s relationship to media and to the people they love — a premium on authentic connection and authentic human moments. But I’m sure the entire distribution will exist.

And the AI glasses are core to that vision.

That’s right.

Please make the case: Why do we need AR or AI glasses when we could just use our phone?

Phones are great — I love phones, I have two of them, they’re wonderful devices. From the very beginning, the question we asked ourselves with glasses was exactly that: phones are great, what’s something you wish you could access without having to take your phone out of your pocket? We came up with a camera and audio — very simple. AI has been a tremendous tailwind, because it unlocks a much larger swath of potential capability over time than what the phone can do through a Bluetooth connection alone. So it’s much more promising now than it looked two or three years ago. Back then it looked like, at some point you’d have to put a display on this and it would have to become a standalone system with a bunch of accessories. Now it looks like there’s plenty of room in the market for a big range of wearable devices — glasses, certainly, and probably other things too. Some of those devices will just be input and output for your phone — that’s cool, your phone’s great, and if it’s making your life more efficient in terms of input and output, that’s awesome. Some will be more complete. For the Meta Ray-Ban Display glasses, for example, we just launched a vibe-coded platform — anyone who wants to can build whatever app they want for the glasses. Right now you build the app and put it on the glasses, but in the future there’s no reason that couldn’t happen in real time, with you telling the glasses what app you want and having it build that app for you on the fly.

I think we’re headed toward a zone that’s a little less app-garden-specific. You’ll still have content homes — content continues to be an evergreen, important thing, as it’s been on TV, on social media, everywhere — so there will still be places where the media you want to reach lives, and those look kind of like apps or channels. But there’s a long tail of things — why does my toaster need an app?

I don’t think it needs one.

I don’t want it. I just want to tell my AI agent, get me the toast I want — it’s the same toast I have every day, just get it for me. I don’t want to go through whatever the steps are.

What does your toaster app do?

I honestly refuse to install it. I absolutely won’t do it

You have to stand up for something.

There’s a line.

It’s sometimes so cool that you can have a specific app to control every aspect of a thing, and I respect that — I’m a tech guy, I like the fidgety nature of it — but it’s kind of gotten out of hand. What I really want is to tell an intelligent system, get me the thing I want, and have it do that for me. We see an early form of this with our Spotify partnership — you ask the glasses to play music, and if you have a Spotify account linked, it goes and gets the music you want. That’s what I wanted — I didn’t want to go through a bunch of steps. So the way I’m thinking about this isn’t that phones aren’t great — they are, and they’ll continue to be — but I don’t think the app model is the way the future looks. I think the future is valuable services provided to you, and you getting access to those services the way you want, when you need them, paying the people who provide them, negotiated either in advance or on demand. I really believe in this.

I noticed the Meta AI app now has a Garmin connector. I’d love to just say, “find me a 5K in my area this weekend and sign me up” —

Totally.

— and do it while I’m on a run, instead of spending an hour figuring it out myself.

Agreed completely. And taking it to a higher level — your Meta AI would ideally already know you’re training, that you have a goal you’re trying to reach, and it’s tied into all the pieces that matter — your nutrition, and so on. That’s the direction we want to take this. There are a lot of steps between now and then, but that’s where we’re going.

Orion and the Future of Full AR

Where do the Orion glasses stand? Those are the full AR glasses we talked about last time.

Orion was such an important moment for us. Having had this AR vision for such a long time, it finally gave us a device we could use to start playing with the software. Even though we couldn’t get the price to a point where we felt comfortable launching it as a consumer product, we designed and developed it with consumer intent — it’s quite wearable, quite workable. I have a pair at home, and we use it to test software. We’ve continued to iterate on the software, and we’ve made a lot more progress — not just because AI has gotten better, though that makes a huge difference, but also because having Orion to develop on makes a big difference.

We continue to be focused on the entire spectrum. We’ve hinted that in addition to display glasses and camera glasses, there’s a whole range of glasses that may sit below that in price. I still believe in full AR as a future for the space. We’re going to continue the same approach we’ve taken so far, and the same reason we didn’t launch Orion — it’s not enough that it does all the functionality, it has to look great, be comfortable enough that you want to wear it, and be at a price point where a reasonable person would say, yeah, this is a good value.

How far away is that?

I’m not going to give an exact number — I’ll say I like the progress we’re making.

Measured in years or months?

I’m not going to answer that.

That’s fair — I had to ask.

I appreciate the hustle.

I know some of my reticence — people who’ve been at companies like ours know this — we’re constantly looking at vehicles and asking ourselves, is this the one, is it ready yet? And, man, we’re getting into the zone. It’s pretty exciting.

Inside Meta’s AI Culture

Let’s talk about Meta culture. You’re running this applied AI division —

I run the Agentic Transformation Accelerator.

Right — which has been the subject of some reporting.

One of the groups in that is the applied AI team.

A Wired source called it “literally the gulag” — zero purpose, barely interacting with anyone, just tasks every week. What’s going on there?

Read more

An epic World Cup has been marred by an overreliance on technology and automation.

In this year’s competition, FIFA has used cameras, sensors, AI systems, and other automated tools to make the game more precise. But these tools, and the accompanying video assistant referee, have taken the soul out of the game.

FIFA’s tech tools have tipped the scales over…

Read more

Read more

There’s a flawed yet oft-repeated narrative that AI adoption in the workplace will inevitably lead to layoffs and job loss. The narrative’s been propagated by some of the most visible leaders in tech, including Anthropic CEO Dario Amodei and Block CEO Jack Dorsey, the latter of whom used AI as a rationale for laying off 40% of his staff.

But a new study …

Read more

Don’t expect AI’s spending frenzy to slow down anytime soon. That was perhaps the biggest lesson at the Big Technology AI Summit, held last week in San Francisco.

Within a span of a few hours, we held a month’s worth of discussions, and the consensus around spending — between infrastructure and services — was that it was only going up from here, even amid the “tokenmaxxing reckoning.”

The summit’s speakers included OpenAI President Greg Brockman, Anthropic Labs Lead Mike Krieger, Box CEO Aaron Levie, Corridor chief product officer Alex Stamos, and several other friends of the shows. Here are the day’s most resonant takeaways, including a look into the state of AI, from Fable and beyond:

Don’t Wait For Moore’s Law

The traditional “rule” that the cost of computing services comes down over time seems broken with AI. The reason is that the technology’s increasing capabilities have led companies to throw it at harder, costlier problems, said Box CEO Levie. “People get confused. They say, ‘I thought AI was supposed to be getting cheaper,’” he said. “Actually, no, we’re outrunning the efficiency improvements in our appetite for what these models can go and do.”

Levie said Box’s token consumption went up exponentially in recent years. “When we launched our first AI use case within Box, the average number of tokens that was being used on a task was like 5,000, 10,000, 20,000 tokens,” he said. “Now our latest agents might use a million tokens or 5 million tokens on executing a task.”

Current AI Policy: The Dream for The Pause AI Movement

Remember the six-month AI pause petition that went nowhere? Levie said that the current AI policy in the U.S. — where the White House seems to be reviewing and approving each new model — is effectively fulfilling the dream of the signatories. “If you had to establish a regulatory regime that said we are going to review models,” Levie said, “you would kind of need something that shocks the system into that kind of regulatory framework.”

Frontier AI Cybersecurity Capabilities: Not Just a Fable Issue

The White House put export controls on Anthropic’s latest model, but Alex Stamos, the former Meta security chief and current Corridor chief product officer, argued the real moment passed without much notice. The AI cyber breakthrough came last year with the Opus 4 and GPT-5 series models, which effectively gave human engineers superpowers. “It’s like all of a sudden you went to a high school track meet and all the kids are running Olympic times,” Stamos said. Mythos, he said, is the best public model at finding bugs. It’s just not unprecedented.

Fable Gets Things Done With Less Back And Forth

Mike Krieger, the Instagram co-founder and current Anthropic Labs lead, said Fable requires less hand-holding than its predecessors. With Fable, he said, he could hand off larger, more open-ended tasks and step away, rather than babysitting the model and going back and forth throughout. “It would do the whole thing, and then kind of hang out for the next seven hours,” Krieger said, of one project.

Compute May Be The Core Differentiator

Greg Brockman made clear that OpenAI believes it will differentiate itself, in part, by the amount of compute it accumulates. “There just is not going to be enough compute in the world to satisfy all the demand,” he said. “Right now, we’re talking about compute constraints and the number of people using these agents is — the order of 10 million, 20 million maybe — we’re not at planet scale.”

Subscribe now

Is Unstructured Data The Key To Successful AI Deployments? (Sponsor)

Live from Hyland CommunityLIVE 2026, Hyland CEO, Jitesh S. Ghai, Chief Product Officer, Michael Campbell and Partha Srinivasa, CIO at Erie Insurance join the Big Technology Podcast to talk enterprise AI, and why unstructured data is the key to unlocking the content-powered agentic enterprise. Where AI pilots and token allotment have gone wrong for others, Hyland is helping its customers—like Erie Insurance—use their unstructured data to provide agents with the context they need to drive real business impact. Tune in to hear how Hyland enables AI deployments in healthcare, banking, insurance, and more by making unstructured data a core part of an agentic enterprise.

Watch Now

What Else I’m Reading, Etc.

Apple will raise prices after memory prices increase [WSJ]

OpenAI may delay its IPO until 2027 [New York Times]

The U.S. is most pessimistic about AI because of its lack of safety net [New York Times]

SpaceX stock tumbles, may drop below its IPO price [Yahoo Finance]

Rest in peace, Om Malik [Om]

This Week On Big Technology Podcast: Anthropic’s Labs Lead On Fable’s Capabilities + Building AI-Native Products — With Mike Krieger

Mike Krieger is the head of Anthropic Labs and co-founder of Instagram. Krieger joins Big Technology Podcast live from the Big Technology AI Summit to discuss what it’s like inside Anthropic the week the government forced the company to pull its frontier models, Fable and Mythos, off the market. Tune in to hear Krieger describe how working with Fable changed the way he builds — queuing up a full night of work before bed and waking to find it finished in an hour — why he insists Anthropic’s safety warnings are material rather than marketing, and how Anthropic navigates being both a platform and a product as it competes with the companies building on top of it. Wired senior correspondent Lauren Goode joins as a co-interviewer. Hit play for a rare look inside the lab from the person building Anthropic’s next breakout product.

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… Moore’s law is still working there.

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

If AI model capabilities continue to improve fast — and relatively in sync across frontier labs — the lab with the most compute will win in the end, OpenAI president Greg Brockman suggested at the Big Technology AI Summit on Thursday.

“There just is not going to be enough compute in the world to satisfy all the demand,” Brockman said. “Right now, we’re t…

Read more

On its first full day of trading, SpaceX is building on its massive IPO’s momentum. The company is up more than 14% on Monday, hitting a market cap above $2.4 trillion, which places it firmly within the ten largest companies in the world.

Now comes the hard part.

To maintain its momentum and lofty valuation, SpaceX will have to turn projections into real …

Read more

When you dig into the numbers, SpaceX, in truth, is a connectivity company with an unprofitable rocket arm and an AI question mark. Elon Musk’s crown jewel IPO’d at a $1.77 trillion valuation today and rose to $2.1 trillion at close, a historic number that’s far greater than its books warrant.

The company lost nearly $5 billion last year after tossing money at AI infrastructure it still doesn’t completely know how to use. It’s been in business since 2002 and still isn’t making money on its rocket launch business. And yet, it just completed the biggest public offering of all time. By far.

Though some may reasonably argue the opposite, the level of support behind SpaceX may be a feature of our system, not a bug. Investors’ willingness to throw money at the seemingly impossible, or crazy, and wait patiently for results is the thing that allows the unthinkable to turn real. It enables investment in improbable initiatives like satellite internet and space data centers, and sometimes those bets pay off. Yes, there’s waste. But to expect perfect efficiency in the moonshot — or Marsshot — business is fantasy.

A similar dynamic is at play in the AI buildout. For all the debate over whether it’s a bubble or underbuild, it’s certain that at least some of today’s record-breaking capital expenditures will be written down, zeroed out, or sold off. We’re already seeing evidence of a coming price war among frontier models, major data center initiatives going sideways, and even SpaceX resorting to selling its compute to rivals (while building an unexpected AI cloud in the process). Despite the drawbacks, AI probably wouldn’t be advancing at its current rate without the financial overexuberance.

One must still pick their bets wisely. Everyday investors are increasingly assuming much of the potential downside if things go wrong. Retail investors accounted for somewhere in the low-20% range of the SpaceX IPO, much higher than the typical 5-10%. And they’ve already put in more than $100 billion of orders for the stock. That amounts to a lot of belief from people placing their hard-earned dollars in a company that didn’t have an AI play a year ago, and now says AI accounts for $26.5 trillion of its $28.5 trillion total addressable market.

SpaceX stock will also show up in a number of index funds, where even passive investors will own it. But actively putting your savings into SpaceX comes with a higher level of risk, perhaps more akin to venture capital than traditional public market investing.

That risk may be the point. It gives companies room to pursue low-probability breakthroughs, even if they may, at times, incinerate some cash on the way there. So SpaceX’s valuation may indeed be a bit out there, but perhaps not as irrational as the numbers alone suggest.

Subscribe now

Mercari built their own JIT access system, then fossilized it and turned to Opal for what’s next (Sponsor)

Maintaining their in-house tool got too expensive as dev tooling evolved, so they migrated to Opal: same zero-touch production posture, zero KTLO.

But the real reason it future-proofs them? AI agents are next.

Now every AI agent is governed like a human identity: least agency enforced, circuit breakers in place, rogue agents isolated.

One identity model for humans and agents.

Learn More

This Week On Big Technology Podcast: What Should An AI Device Look Like? — With Alex Himel

Alex Himel is Meta’s VP of Wearables. Himel joins Big Technology Podcast to discuss the future of AI wearables and why Meta believes glasses could become the next major computing platform. Tune in to hear how AI assistants might help with daily tasks, meetings, reminders, fitness, photos, and real-world context without pulling people out of the moment. We also cover the competition from OpenAI, Google, Apple, and Amazon; the privacy questions around facial recognition; on-device AI; and the story of how Mark Zuckerberg pushed Meta to turn Ray-Ban glasses into an AI product. Hit play for a sharp look at whether smart glasses are finally ready to break through, and what they could mean for the future of personal AI.

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 less risky than a Marsshot!

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

When Tim Cook takes the stage at Apple’s developer conference in Cupertino today, he’ll be leading Apple through yet another moment of transition.

It will be Cook’s last WWDC as CEO as he makes way for incoming leader John Ternus. And it’ll be the second WWDC in three years to promise AI improvements to Siri and iOS as Apple transitions into its artific…

Read more