A (0:01)

Today on the AI Daily Brief. Who cares about consumer AI? Before that in the headlines were Coinbase's layoffs really about AI? The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. Alright friends, quick announcements before we dive in. First of all, thank you to today's sponsors, kpmg, Granola Mercury and Section. To get an ad free version of the show go to patreon.com aidaily brief or you can subscribe on Apple Podcasts. Remember, ad free is just $3 a month. If you want to learn more about sponsoring the show or really anything else about the show or around the show, go to aidailybrief AI. Quick reminder that on aidailybrief AI we are live with the AI Usage Pulse Survey. Thanks in advance for taking a minute to do that. You can find a link in the show notes as well and you can also find links to our latest education programs. Join 4,500 friends to learn how to build your agentic operating system with the Free Agent OS program. Or if you're looking for something for your teams, we've got the third cohort of Enterprise Cloud Live as well as a new four week executive Sprint from New Far Gaspar that is registering people now as well. Again, you can find all that on aidailybrief AI. Earlier this week, as you well remember, I made a big deal about a bunch of fairly small signals that suggested to me in aggregate that maybe, just maybe the AI doom narrative was feeling a bit of narrative exhaustion, that maybe there was a opening for a more optimistic narrative surrounding AI. The response to Coinbase's layoff announcement yesterday suggests that maybe I was overstating the case. Yesterday, Brian Armstrong tweeted the email that he had sent to all employees at coinbase announcing a 14% reduction in the team. That means about 700 of 5,000 people will be laid off. And of course I was at the very center of the story. In fact, if you just look at the news headlines about this, AI was the only part of the story. Reuters Crypto Exchange Coinbase to cut about 14% of workforce in AI driven cuts Fortune Coinbase didn't just lay off 14% of its staff due to AI. It replaced managers with player coaches and turned its blah blah blah blah blah. SFGate, formerly SF headquartered crypto giant to lay off 700 workers in AI shift. And of course the New York Times, the paper of record. Coinbase lays off 14% of employees as AI changes work so what's the issue with this? Brian did say explicitly, AI is changing how we work. He wrote. Over the past year, I've watched engineers use AI to ship in days, what used to take a team weeks. Non technical teams are now shipping production code and many of our workflows are being automated. The pace of what's possible with a small focus team has changed dramatically and it's accelerating every day. All of this has led us to an inflection point, not just for Coinbase, but for every company. The biggest risk now is not taking action. We are adjusting early and deliberately to rebuild Coinbase to be lean, fast and AI native. We need to return to the speed and focus of our startup founding with AI at our core. The rest of the post is all about what that means. Fewer layers, faster decisions, no pure managers. Every leader at Coinbase, he says, must be a strong and active individual contributor. Managers should be like player coaches, getting their hands dirty alongside their teams. He also announced what he called AI native pods. We'll be concentrating around AI native talent who can manage fleets of agents to drive outsized impact. We'll also be experimenting with reduced pod sizes, including one person teams with engineers, designers and product managers all in one role. In short, AI is bringing a profound shift in how companies operate and we're reshaping Coinbase to lead in this new era. This is a new way of working and we need to leverage AI across every facet of our jobs. So, first of all, I don't not believe any of what he's saying there. In fact, a lot of that is the type of stuff that we talk about on this show every day. What's more, Brian has a long history of making tough decisions with Coinbase that have helped it weathered and be a strong company across what is one of the most volatile industries in the world. So what's the issue? Well, the issue is, of course, with that volatile industry that Coinbase is in, as much as they are trying to move into other areas of finance, Coinbase is, and remains at the moment, a crypto company. Crypto is, to put it mildly, not doing well. It's beyond the scope of the show to get into whether that is regular cyclicality, much more concerning fundamentals. But there are, I will say, plenty of people, including many of the big financial giants who made a big deal about getting into the space 18 to 24 months ago, who have lost the faith. And to be fair to Brian, he did spend at least a couple sentences on crypto, although it's pretty minimal. Crypto, he writes, is on the verge of the next wave of adoption with stablecoins, prediction markets, tokenization and more taking off. However, our business is still volatile from quarter to quarter. While we've managed through that cyclicality many times before and come out stronger on the other side, we're currently in a down market and need to adjust our cost structure now so that we emerge from this period leaner, faster and more efficient for our next phase of growth. What he's not saying is holy crap. Absolutely no one is trading crypto anymore as evidence of that. Just a week ago when Robinhood announced their Q1 earnings, they showed a 47% year over year dip in crypto trading revenue. So if you are a public market CEO and your option is either a talk about how your core market is halved year over year, which I'm not saying Coinbase's numbers are the same as Robinhood's, I'm just using it as an example and opinions about that industry are incredibly bleak. Or alternatively, are you going to blame it on AI? It's not even a question and I'm not even mad at Brian Armstrong for doing it. What's preposterous is the unanimous wall to wall acceptance of AI as the main story. It's almost like media outlets will question CEOs unless those CEOs happen to be reinforcing their preferred narratives. Bucocapital on Twitter wrote, the only companies firing people because AI makes them so wildly productive also share these attributes. Overhire during COVID Our market share losers have giant capex spend what a delightfully curious coincidence. Axios was alone in questioning the narrative and good on them. They published a piece called AI Becomes the Easy Alibi for Waves of Layoffs. Look, I will end the rant there, but my point is ultimately this. There are absolutely going to be massive changes to the structure of employment, certain types of roles, entire industries. Change will come because of AI. I want us to deal with the real stuff rather than just have this be a convenient excuse for public market CEOs looking to defray the bigger story. Moving on to our next story, but staying in markets. Anthropic has committed to spending 200 billion with Google Cloud as new details emerge about their blockbuster deal. Last month Anthropic announced a new deal with Google and broadcom to supply 5 gigawatts of compute starting from next year. No dollar amount was attached, but this is a huge amount of compute, so no one thought it would be cheap. The information now reports the deal is worth 200 billion spread across five years and indeed represents the lion's share of the $462 billion backlog Google announced last week during earnings, which sent their stock soaring. Digging into the story a little further, the information found that Microsoft, Oracle, Google and Amazon have now reported a $2 trillion backlog between them, with OpenAI and Anthropic Accounting for almost half of that backlog. The article of course ran through the usual arguments we've heard over the past year, the question of whether revenue from the Foundation Labs can keep up with their spending commitments, as well as of course, the circular spending argument where in this case Google's investment of up to 40 billion in anthropic seems to be flowing back into their financial reports as backlogged orders. Now, to many, those arguments are getting less and less interesting as we see revenue growth go vertical this year. But what's interesting is how differently investors seem to be responding to this news compared to similar previous reports. Last September, Oracle announced they had a $455 billion backlog, largely consisting of a $300 billion deal with OpenAI spread over five years. The stock jumped by 36% in a day, one of the single largest day moves for a company of their size in history. Yet in the following days, analysts very loudly questioned whether OpenAI could meet such a huge commitment. The stock retraced the entirety of its move by mid November and has been grinding lower ever since. It only began to slightly recover last month. The response to the Google Anthropic deal was the polar opposite. The market already knew Google had a $462 billion backlog last week, but Google's stock surged on the news that Anthropic was behind over 40% of the backlog. The news was released just after market close, with the stock adding 1.5% in overnight trading after an already strong session which saw it reach a new all time high. The boost was enough in overnight trading to push Google up above Nvidia as the most valuable company in the world, although that has not persisted into trading today. So maybe that idea that the narrative is shifting isn't completely wrong. Palantir also had a massive earnings session Heading into the report, the big question was whether Palantir should be grouped with Big Tech as an AI beneficiary or be lumped in with the software slowdown. Palantir stock is down almost 20% so far this year, but their AI first platform positions them to benefit from the token drought. The big number was top line revenue, with Palantir reporting 85% year over year growth, their fastest pace since their public market debut in 2020. Net income is up 4x over the past year reaching 870 million for the quarter, with government based revenue driving the acceleration increasing from a 66% growth rate in Q4 to 84% in Q1. CTO Shyam Sankar had some very clear thoughts about Palantir's position in the rapidly emerging token economy. Tokens are the new coal, he said. Palantir is the train. Speaking of coal and trains and core commodities, BlackRock CEO Larry Fink said that he believes that COMPUTE will become a financialized commodity. In a panel discussion at Milken on Tuesday, Fink said that he thinks AI COMPUTE will eventually be traded on futures markets similar to oil, wheat or electricity. He said that a new asset class will be buying futures of compute. Fink also informed his audience of financial industry folks that the US Is tapped out on capacity, commenting, we just don't have enough COMPUTE power right now. Putting a fine point on his current view of the industry, he added, there is not an AI bubble. There is the opposite. We're short power, we're short compute, we're short chips. Demand is growing much faster than anyone has ever anticipated. Fink also teased that BlackRock is preparing to announce an infrastructure funding deal with the Hyperscaler in the coming weeks, putting a fine point at least on the market's return to enthusiasm. Demand for the Cerebras IPO is massively outstripping supply. Earlier in the week, SEC filing stated that the firm plans to sell 3.5 billion at a high end valuation of 26.6 billion. The problem, according to Bloomberg sources, is that private investors are seeking 10 billion in allocations. On Tuesday, Bloomberg reported that the presale is turning into an auction. Investors have been told to submit their desired allocation and maximum price. This is a break from typical IPO protocol. Usually the price is set and excess demand is resolved by trimming allocation. This unusual practice could suggest that Cerebrus will raise their IPO price and maybe bump up the number of available shares. It also is very clearly an indicator on just how bullish Wall street is on AI chips right now, which I think we'll get confirmation of once the IPO goes live. For now though, that is going to do it for the headlines. Next up, the main episode. One of the most important AI questions right now isn't who's using AI, it's who's using it? Well, KPMG and the University of Texas at Austin just analyzed 1.4 million real workplace AI interactions and found something surprising. The highest impact users aren't better prompt engineers. They treat AI like a reasoning partner. They frame problems, guide thinking, iterate, and push for better answers. And the good news? These behaviors are teachable at scale. If you're trying to move from AI access to real capability, KPMG's research on sophisticated AI collaboration is worth your time. Learn more at kpmg.com us sophisticated that's kpmg.com us sophisticated Today's episode is brought to you by Granola. Granola is the AI notepad for people in back to back meetings. You've probably heard people raving about Granola. It's just one of those products that people love to talk about. I myself have been using Granola for well over a year now and honestly, it's one of the tools that changed the way I work. Granola takes meeting notes for you without any intrusive bots joining your calls. During or after the call, you can chat with your notes, ask Granola to pull out action items, help you negotiate, write a follow up email, or even coach you using recipes which are pre made prompts. Once you try it on a first meeting, it's hard to go without. Head to Granola AI AIDAutaily and use code AIDAutaily. New users get 100% off for the first three months. Again, that's Granola AI AIDAutaily. This episode is brought to you by Mercury Banking for people who expect more from the tools they rely on. If you're building a modern business but still using a traditional bank, it just doesn't make sense. I use Mercury for all of my ADB family of companies and it honestly feels like financial software built for how people actually operate. Today. It's fast, clean, no in person visits, no minimum balances, and the things that used to take forever like sending wires or spinning up new accounts take seconds. Everything lives in one dashboard, cards, payments, invoices, team permissions. And you can automate a lot of the busy work so you're not constantly manually managing your money. Of all of the services I use to run aidb, I never thought banking would be one of my most painless and most happy experiences. But with Mercury, that's exactly what it is. Visit mercury.com to learn more and apply online in minutes. Mercury is a fintech company, not an FDIC insured bank. Banking services provided through Choice Financial Group and Column NA members FDIC Here's a harsh truth. Your company is probably spending thousands or millions of dollars on AI tools that are being massively underutilized. Half of companies have AI tools but only 12% use them for business value. Most employees are still using AI to summarize meeting notes if you're the one responsible for AI adoption at your company, you need section. Section is a platform that helps you manage AI transformation across your entire organization. It coaches employees on real use cases, tracks who's using AI for business impact, and shows you exactly where AI is and isn't creating value. The result. You go from rolling out tools to driving measurable AI value. Your employees move from meeting summaries to solving actual business problems and you can prove the roi. Stop guessing if your AI investment is working. Check out section@section AI.com that's s e c t I o n a I.com welcome back to the AI Daily Brief. Today we are discussing who cares about consumer AI, which in this case is both a question and a statement. We've got a couple of pieces of news, plus some rumors and reports that all dealt to some extent with consumer AI. And it almost served as a reminder of just how much the focus of the industry has moved away from the consumer side of things and centered squarely on the enterprise. Now to understand where this came from, we gotta go back in time a little bit. Q4 2025 was a rough one for OpenAI. The reception of GPT5 had been quite bad. Google seemed to be firing on all cylinders with all of their Gemini products winning more and more users, and the devotion of developers was almost entirely in the anthropic camp and had been at that point for about a year and a half. Now, even as early as GPT5, it was clear that OpenAI was trying to shift the conversation around it, that OpenAI was at least trying to nudge the conversation towards use cases like coding as opposed to just the consumer usage where they were dominant. The company started talking about codecs a lot more. Indeed, they were releasing codec specific versions of each of their incremental model releases. Fast forward a couple months from there, the world is getting clawified. OpenAI has actually brought the creator of OpenClaw in house and CEO of Applications Fiji Simo is pushing the company incredibly hard to cut back on side quests to focus on their core business. Now it was clear even at the time that by core business she meant the coding and enterprise business, and by side quest she meant everything else. But what made that real in a way that basically nothing else could was when OpenAI went so far as to actually shutter their Sora app. This was one of the first times that we had seen OpenAI actually have to make a choice in where to deploy its compute. And by canceling this consumer video app and the billion dollar Disney deal that came with it, they could not have been sending a clearer signal that enterprise and work related use cases were, for the foreseeable future, their focus. And now we're operating in a world where the entire narrative of 2026 has been all about not only coding agents, but coding agents and agent decoding processes implement, impacting every other knowledge, workflow and just a huge amount of the attention being squarely on the enterprise. Airbnb CEO Brian Chesky said that of the 175 companies in the last batch of Y Combinator, only 16 weren't focused on the enterprise. And yet, even if OpenAI is still young enough to get to decide what its core DNA is, there are some companies that are just unavoidably consumer through and through. The most obvious of these is Meta. And Zuckerberg, for his part, seems hell bent on building consumer AI, even if it looks like them swimming against the tide. Which is not to say that they're not inspired by what's going on in enterprise AI. The information reports that Meta is training a new Open Claw inspired agent that is codenamed Hatch. Sources said that they aim to have the product ready for internal testing by June. The agent is designed to be the logical progression of many of the early tests in consumer Agentix, largely focused on shopping and personal productivity tasks. Meta is currently training the agent to navigate the web in simulations of real websites and apps using DoorDash, Etsy, Reddit, Yelp and Outlook. Currently, Hatch is powered by Claude models, but Meta intends to use their own models upon release. In addition, Sources said that Meta plans to integrate a separate shopping agent into Instagram with a target launch date of Q4. And increasingly, meta is alone as the only major lab that has consumer AI as their current and dominant focus. During last week's earnings call, Mark Zuckerberg said that Meta's goal is to, quote deliver agents that can understand your goals and then work day and night to help you achieve them. But that does not seem to be about your work goals. In that same earnings call, Zuckerberg moved the conversation away from the coding agents that have dominated the B2B conversation. He said, I'm not against having an API or coding tools, but it's not our primary focus. People conflate coding with self improvement more than they should. Coding is one ingredient for the model. Self improving. It's not the only thing we are focused on. All the parts that are going to be necessary for self improvement. Now what makes Zuckerberg's bet all the more interesting was is that they are forecasting to spend between 125 billion to 145 billion in infrastructure spend this year, meaning that they clearly believe that there is financial opportunity that others aren't seeing in the consumer space. Another story from yesterday that is at least nominally consumer related, although not exclusively OpenAI has updated their default model with the release of GPT 5.5 Instant. It replaces GPT 5.3 Instant, which even after the release of GPT5.5 thinking, was the default choice for Free users as well as subscribers on the $8 ChatGPT Go plan. And as OpenAI has moved away from UI emphasis on the model selector, the choice of their default model has more impact on the consumer experience. Indeed, with the introduction of GPT5.3 Instant in March, OpenAI actually removed the model selector for Free and Go users. Now, 5.5 Instant itself is a good model. It delivers a significant benchmark jump over its predecessor. It scored 81.2 on the AIME 2025 math test compared to 65.4 for 5.3, and the multimodal reasoning benchmark MMLU Pro also saw a jump from 69.2 to 76. Probably even more importantly, OpenAI says they've seen reduced hallucination rates for sensitive areas, and the model will now be able to access memory, use a Gmail connector, and have better context management. Ethan Malik noted that the Benchmark suggests that OpenAI's free model is now at a similar level to Frontier models from late last year. Now, part of why this matters is that one of the big divides in the AI conversation has been that for some time the skeptics have been able to convince themselves that AI actually isn't all that good. This has finally started to go by the wayside, but a lot of the loudest critical voices were peddling this idea that people didn't really have to pay attention because the industry was just hyping itself up. And while anyone who has used, for example, CLAUDE code to build something might be gobsmacked at that assessment, you have to remember that for about 900 million people, their only experience with AI is whatever the base model that ChatGPT gives them. If 5.5instant proves to be as big a jump in quality as it seems like it could be, it might dramatically change what those people think of AI, helping some of them finally reassess their priors. And yet it still certainly feels like an afterthought compared to the enterprise focused and coding focused model releases the questions surrounding consumer versus Enterprise AI aren't just limited to the industry itself. During an appearance on Tuesday alongside anthropic star AO Amade, JP Morgan CEO Jamie Dimon endorsed the AI capex boom, but said that he wasn't sure that this was actually a consumer technology. He said the technology is so powerful it's worth the trillion dollar investment, and argued that enterprise use cases have found their niche. On the flip side, when it comes to consumer, he said, it's not clear to me how consumer is going to play out. A lot of you probably use Gemini, you can use it for free and that may be completely sufficient for your requirements. Enterprise we always look at it a little bit more like I'm making an investment and what I'm getting for it. If it makes you better off, you make the investment, whether it's hardware, software or machine learning or salespeople. The point that he was making was not that consumer AI was useless or anything like that, but just that in his estimation, there wasn't an obvious basic consumer use case that would warrant a paid subscription. Again, outside of work. Another interesting example to see just how much things have changed is looking at the responses to last year's GPT image model release versus this year's in 2025, the release of nanobanana. But even more GPT images were some of the biggest viral moments of the year. To this day, there are still people using a Studio Ghibli version of their portrait as their profile picture and giblifying memes. According to a recent study from App Figures, these releases were by far the biggest download drivers for the Gemini app and the ChatGPT app respectively. Last year, Nanobanana drove 22 million additional downloads in a month, while GPT images added 12 million incremental downloads. Those two models were the second and third ranked consumer AI releases, falling short only to Deepseek R1, which as we've discussed extensively was the first time most people ever got their hands on a reasoning model. There was also just a tremendous amount of buzz and horror frankly from the markets around that, leading more eyes to be on it. The App Figures report also highlighted that GPT Images was a much larger event than any other text based model OpenAI had released. Compare that to the release of GPT Images 2 this year. It didn't generate anywhere near as much hype. The only meme y type viral moment we got was replicating a five year old's drawing in Ms. Paint that was popular enough to be a couple of covers for my shows last week, but has already faded. Now part of this might be simply the fact that images too are although massively better on vectors like text and editability didn't bring a net new experience to the consumer. It was a major upgrade for people that use image generation as a tool, but for the casual user it didn't represent something fundamentally new. And in fact, when you actually look at the discourse around images too, it was basically all about how people could use it with codecs. To put it bluntly, if last year's big drivers were image generation models, this year is all about the coding and work harnesses. People, for example, have been following the dogfight between Codex and Claude code, with Ticker trend showing a huge increase in Codex installs. Simon Smith points out that the value of these new model releases is basically just about how they work in the harness together, sharing Codex zooming ahead of Claude code and download Simon Smith writes the Frontier AI competition is ongoing between primarily Google, OpenAI and anthropic right now, but this shift in preference to Codex is a huge recent change, driven I think largely by the triple Whammy of GPT Image 2, GPT 5.5 and Codex all being fantastic and working well together. Now what he means and what I mean when we're talking about Image two being just part and parcel of this larger Codex conversation is that the thing that people have been so excited about is using GPT Image 2 to solve the UI and design issues that had previously hampered OpenAI models as compared to Anthropic models. The thing I think that we haven't quite named as crisply as we need to to understand what's changing is that up until this point, the last three to six months, one could reasonably look at attention on consumer AI versus attention on enterprise AI as a question of seat conversion. Total pie of consumer is bigger, so even though they convert to paid seats at a lower rate, they might be worth more in aggregate. That seemed to bear out by how much bigger OpenAI was in terms of revenue than Anthropic throughout 2025. But then Anthropic went on this massive tear in 2026. It surged from 14 to then 19 to then 24 to 30 and now apparently to 44 billion in annualized revenue. And the important thing is that that is not because they uncovered a whole bunch of new seats in the enterprise that they could convert to paid users. It's because work related usage of these models is categorically different than the seat based consumer usage. An individual using the API in Claude coder codecs is not potentially worth 10 times as much as a user. It's not the difference between a $20 subscription and a $200 subscription. They are worth potentially 100 or even more times that user. In fact, it's not even clear yet that we really actually have a sense of just how much more an API user is worth relative to a normal seat user. What we do know is that the Labs are having to shift their business models to move towards consumption based usage because the power users are consuming so much. What's more, I don't think anyone thinks we've seen even close to the cap on the demand for tokens. And in a world where the demand for tokens from work related users massively outstrips the supply, how can it possibly make sense to reserve even a handful of those tokens for the consumers? Certainly people are noticing that the level of products is not nearly the same in these two areas. A Reddit conversation in R AGI from a few months ago was why are AI coding agents incredible while consumer agents suck? Now it is worth noting that not only has consumer AI not failed, but it is by far unquestionably the fastest growing technology category in history. The total weekly active users of AI has gone from about 100 million at the beginning of 2024 to 1.2 billion in 2026. You're talking about a 20x plus increase in two years, as investor Apoor Vagarwal puts it. ChatGPT at 900 million weekly active users is larger than Spotify at 600 million, in the same neighborhood as TikTok and Instagram, and approaching WhatsApp and Chrome, each of which I consider essential utilities. ChatGPT used to be a rounding error next to consumer giants. Now it's in the same league. Most people, myself included, thought that this would take much longer. And to be clear, these aren't just curiosity users when it comes to metrics of engagement like the ratio of weekly to monthly active users, basically showing that most people are using it frequently rather than just irregularly. ChatGPT at this point is ahead of X Spotify and TikTok and starting to come up into Facebook and Instagram territory. ChatGPT is also getting stickier over time, which very, very rarely happens. The time that individual users spend on AI apps is also rising. ChatGPT has roughly tripled its time per user since early 2023. The point being that this is not a question of whether consumer AI is working, it very clearly is. It's a question of money. The base level analysis is about the very small fraction of users who pay for premium accounts. Bank of America, for example, recently came out with a study that found that only 3% of households where bank of America customers paid for AI today. It's really about the fact that the token consumption value of a work related user has become, as I stated before, categorically different to a consumer user. So how can this possibly resolve? The short answer is there has to be some revenue stream outside of subscriptions that makes consumer worth paying attention to for these companies. Olivia Moore from A16Z writes, the big story that most people are missing in the AI race for the consumer is ads. Right now most consumer AI revenue is coming from power users who are willing to pay high cost subscriptions. This will not be the end state. Google makes around $460 per user per year in the US mostly on ads. Meta makes around $250. I would argue ChatGPT's ad based ARPUs will be even higher as they will ultimately have deeper and more frequent user engagement. But even at the $460 level that is matching Google, currently monetizing everyone in the US via ads is 152 billion in annual revenue. By contrast, if you're able to monetize even 5% of the population on a $200 a month subscription, that's only 40 billion. And that's the logic of why, even if people get up in arms about ads for consumer AI to exist, it feels almost inevitable that they're going to be part of the equation. And while the announcements get buried relative to everything else, OpenAI does keep adding developments around their ad platform. Another area that some have looked at is agentic commerce. Although there's even more skepticism here, I think that there is around ads. Ads at least are proven territory. Whether people want agents to shop for them I think is very different. Andy Jassy from Amazon recently talked about this. What you see with agentic commerce, he said, is it is a small fraction of what we see with the search engine referrals. But the experience has just not gotten great with these third party horizontal agents yet. They're not often able to get the pricing right or the product information right. They don't have any personalization data or any shopping history. So we do want to see that get better with third party horizontal agents. We're having conversation with all those folks to try and make that better and find something that works for consumers in all the companies. It will be interesting over time which agents customers choose to use. Jassy's argument is that most people will default to the native agentic shopping assistant in the merchants they already shop with. I think personally that one of the challenges is that shopping represents two very different types of behavior. There is the version of shopping where people just want the thing that they want and they don't want to think about it very much. And for that, yes, potentially agents could be really good in a lot of shopping. However, the browsing is part of the experience. People like the quest. Learning about all their options is part of the value they get. It's less clear to me how agents fit in with that. And by the way, even in the first scenario, I think people might be underestimating the cognitive cost of offloading all the context that a shopping agent needs to actually make the right decision for you. Are people going to sit down and tell agents all of the criteria they have only to realize when the results come back that they left off something really important that they hadn't even thought of before? At that point, why aren't you just doing it yourself now? I'm not at all saying that these things can't be solved. I'm just saying it's another big challenge that makes consumer AI look tough. Devices could be another area where consumer AI flourishes. Leaks are suggesting that OpenAI is accelerating development of its first AI agent phone, with mass production potentially starting in the first part of next year. But there are also reports that they're thinking about spinning out their robotics and hardware divisions to have more focus. When it comes to the question of who cares about consumer AI, there is no question that a lot of consumers care about consumer AI. ChatGpting has already started to become a verb in the way that googling has been for 20 years. But the challenge is that we are entering a period of scarcity where, more than it's ever been, the supply of tokens will not be able to keep up with the demand of tokens in that period, consumer AI looks very tough to me. But then again, that might be why it's an interesting contrarian bet. Back to that same Brian Chesky interview where he talked about YC going to the enterprise. After describing all the different reasons why companies he didn't think were going after consumer, he summed up maybe finally, the reason people aren't doing consumer companies is that they're just harder. You have to be good at a lot more things you generally have to be better at design, marketing, culture, press. It's not purely technology and sales. But, he continued, my prediction is that we're living in the age of enterprise AI. But I think in the next 12 to 24 months, you're going to see the beginning of a consumer AI renaissance. Almost every app on my home screen has not changed since AI, including Airbnb. I think that's going to change in two years. So in answering who cares about consumer AI if you're standing there raising your hand, maybe it's time to build for now. That's going to do it for the AI Daily Brief. Appreciate you listening or watching as always. And until next time, peace. Sa.