Today on the AI Daily Brief, why every AI app is turning into every AI app. Is it distraction and product confusion or about something more fundamental? Before that of the headlines, NVIDIA CEO Jensen Huang suggests politely that maybe AI leaders could stop scaring the ever-loving poo out of everyone. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. All right, friends, quick announcements before we dive in. First of all, thank you to today's sponsors, KPMG, Blitzy, AIUC, and PromptQL. To get an ad-free version of the show, go to patreon.com slash ai daily brief, or you can subscribe to Apple Podcasts. If you are interested in sponsoring the show, send us a note at sponsors at aidailybrief.ai. And of course, aidailybrief.ai is where you can find out everything that is going on in the AIDB ecosystem as well. Always lots of good stuff cooking there. But let's dive into the headlines. It is a truth universally acknowledged that there has never in the entire history of business communications been any set of people so spectacularly bad at communicating as the contemporary leaders of the AI industry. Really, since the launch of ChatGPT, it has just been a clinic in how not to talk to people and how not to build public support for what you're building. NVIDIA CEO Jensen Huang has finally had enough. Ever since the beginning of Gen AI's rise to prominence, Huang has been nothing but optimistic. He has continuously argued and never moved off his stance that AI is going to create jobs, and he's never given quarter to any sort of AI takeover theories, instead dismissing them as science fiction. Now he's calling on AI leaders to follow his lead. At a panel at the company's GTC event, he said, the desire to warn people about the capability of the technology is really terrific. Warning is good. Scaring is less good, because this technology is too important to us. Now, going farther, Jensen thinks that in the midst of a growing national security debate around AI, Huang believes that one major national security risk is AI pessimism. This is, of course, something that we've talked about extensively on this show, and an area where I very much agree. Americans consistently rank as some of, if not the least optimistic about the technology, which has major implications for everything from adoption to policy and beyond. Huang said that the anger and fear around AI could cause the US to fall behind other nations, and I would go further, it absolutely 100% will. Huang then urged AI leaders to bring the conversation back to what the technology actually is, not the highly speculative discussion of what it could become. He commented, It is not a biological being. It is not alien. It is not conscious. It is computer software. To say things that are quite extreme, quite catastrophic, that there's no evidence of it happening, could be more damaging than people think. Now, of course, reasonable people are going to disagree on the line between warning and thoughtful discourse about possibilities and outright scaring, but it feels pretty clear to me that at least someone needs to take on the job of articulating what the positive future with AI could look like, because that exists basically nowhere in the discourse right now. And of course, things aren't going to get any less controversial from here. Jeff Bezos is in talks to raise a $100 billion fund to transform the manufacturing sector using AI. The Wall Street Journal reports that Bezos has met with some of the largest capital managers in the world over recent months. Sources said he met with sovereign wealth funds across the Middle East earlier in the year, and more recently visited Singapore as part of the effort. Investor documents describe the fund as a manufacturing transformation vehicle. It aims to buy up companies in major industrial sectors, including chip making, defense, and aerospace. The effort is linked to Project Prometheus, which was a startup founded by Bezos last November. That company aims to train AI that understands the physical world for deployment in engineering and manufacturing. Bezos, it would appear, seems to be applying the private equity model of buying out legacy firms and revamping their tech stack to physical industries. The goal, of course, is by developing the technology and buying up the customers to build a massively vertically integrated effort to deploy physical AI at scale. Now, there is an interesting broader shift here, where even as software starts to eat itself as AI forces margins down, more and more entrepreneurs are moving back from bits to atoms and exploring the physical world again. Meanwhile, the politics of this one are already fraught, with Bernie Sanders tweeting, Jeff Bezos, worth $234 billion, plans to replace 600,000 American workers with robots. Now he wants to spend $100 billion to fully automate not just his warehouses, but factories in the US and other countries. Oligarchs are waging all-out war against workers. Fight back. Bernie Sanders also tweeted out a video of him having a conversation with Claude about, as he put it, AI collecting massive amounts of personal data and how that information is being used to violate our privacy rights. This one admittedly was pretty weird, but if you're wondering whether Bernie is going to let this AI stuff go, the answer is clearly no. Speaking of AI policy, the White House is set to announce a legislative framework for federal AI rules. Axios reports that the administration is expected to instruct Congress on their regulatory preferences today, although the details as I record are not yet available. There is some amount of increasing pressure for Congress to get AI regulation on the books heading into the midterms. Over the past year, the administration has been clear in their position that AI regulation should be a federal matter, but there's been a lack of consensus on exactly what the framework should look like. It is, however, increasingly untenable for the administration to resist state regulations without putting forward their own clear set of policy preferences. Earlier this week, OpenAI Chief Global Affairs Officer Chris Lehane threw in his lot with state regulators, writing in a blog post, in the absence of a national framework, states should align around the emerging model in California and New York. Also this week, Google's President of Global Affairs, Kent Walker, welcomed state coordination on AI and called the approaches from California and New York manageable frameworks. According to the Axios reporting, this new federal framework will preempt state regulation and tackle the four C's as previously laid out by AI czar David Sachs. Those topics are child safety, communities, creators, and censorship. Some of these issues are fairly easily resolved. For example, the proposal is expected to codify the president's ratepayer protection pledge, which requires tech companies to pay for their own energy infrastructure. But other issues are very quickly becoming quagmires. On Wednesday, Republican Senator Marsha Blackburn also released her own discussion draft of a bill which she claimed represented the administration's views. That draft included duty of care provision, the ratepayer protection pledge, deepfake protections, and a set of guidelines around content watermarking. Wildly controversially, the draft would sunset Section 230 of the Communications Decency Act, which protects online platforms from liability associated with user-generated content. While many have called for reforms to Section 230, a full repeal is not something that is going to just go through without consideration, given that it's pretty much the foundation of the modern social internet. Despite Republicans' reputation for lighter-touch regulation, Adam Thera writes that Blackburn's massive new AI regulation bill, 291 pages of near-endless mandates, would, quote, make European technocrats blush with envy if it ever passed It represents he says a recipe for technological stagnation and hyper of technology markets and speech that must be completely rejected So yeah If you thought we were close to some common sense rules we are it appears not Lastly today, Apple's App Store is throwing the brakes on the Vibe Coding revolution. And yet many think their rules are out of step with the AI era. The information reports that multiple Vibe Coding platforms, including Replit and Vibe Code, have been blocked from updating their apps unless they make big modifications. The App Store prohibits apps from running code in a way that changes the way the app functions, and that nebulous rule is now being enforced, leading to a crackdown on mobile VibeCoding platforms. An Apple spokesperson said that the policy wasn't specific to VibeCoding apps, and sources added that Apple is close to reaching an agreement with Replit and VibeCode, with each agreeing to either tweak how previews are presented or remove certain features entirely. Replit said their tweaks involve showing previews in a separate browser rather than in the app. VibeCode said that they had been instructed to remove the ability to VibeCode apps for Apple devices entirely. And while the policy is theoretically born out of security concerns, there is an obvious chilling effect that some believe is deeply cynical. Gene Burris, a competition lawyer who works with the Coalition for App Fairness, said, Apple has a history of not allowing apps or features that create competition on their platform. And indeed, others are calling for Apple to get with the times, even if it means consumers can create their own software rather than paying the app store tax. Kyle Maycomber, the CEO of Vibecoating platform BitRig, said, I think vibe coding is really compelling and people want it, and so I hope Apple will notice this and the value it brings and is working on revised guidelines. Maycumber is himself a 14-year Apple veteran before founding his own company, and while he understands the security concerns, he noted that the policies were put in place many years ago. Gauntlets Austin Allred writes, App Store Review was one of the first columns of the software ecosystem to just completely buckle under the weight of AI. It almost makes building apps not worth it until Apple gets its stuff in order. That said, putting his tongue in his own cheek, he wrote, why is the App Store review taking so long? He complained as his agent submitted the five new apps it had built that day to the App Store. This is a problem that is going to absolutely get worse, not better. So Apple's got to do something here, and I don't think broad, blunt policy is really going to work. Vibe coding is, however, in a way, the genesis topic of our main episode as well. So for now, we will close the headlines and move over on Into the Main. All right, folks, quick pause. Here's the uncomfortable truth. If your enterprise AI strategy is we bought some tools, you don't actually have a strategy. KPMG took the harder route and became their own client zero. They embedded AI and agents across the enterprise, how work gets done, how teams collaborate, how decisions move, not as a tech initiative, but as a total operating model shift. And here's the real unlock. That shift raised the ceiling on what people could do. Humans stayed firmly at the center while AI reduced friction, surfaced insight, and accelerated momentum. The outcome was a more capable, more empowered workforce. If you want to understand what that actually looks like in the real world, go to www.kpmg.us slash AI. That's www.kpmg.us slash AI. Blitzy is driving over 5x engineering velocity for large-scale enterprises. A publicly traded insurance provider leveraged Blitzy to build a bespoke payments processing application, an estimated 13-month project, and with Blitzy, the application was completed in live in production in six weeks. A publicly traded vertical SaaS provider used Blitzy to extract services from a 500,000-line monolith without disrupting production 21 times faster than their pre-Blitzy estimates. These aren't experiments. This is how the world's most innovative enterprises are shipping software in 2026. You can hear directly about Blitzy from other Fortune 500 CTOs on the Modern CTO or CIO-classified podcasts. To learn more about how Blitzy can impact your SDLC, book a meeting with an AI solutions consultant at Blitzy.com. That's B-L-I-T-Z-Y dot com. Quick update on something I've been following. AIUC1 is the first real standard for AI agents, developed with Fortune 500 security leaders to basically define what safe, enterprise-ready AI agents should look like. A little while back, I mentioned that 11 labs became certified against AIUC1. This week, two more big players joined, Finn from Intercom and UI path. What that certification means in practice is real-time guardrails that block unsafe responses, protection against manipulation, and a full safety stack designed for enterprise environments. And that's why this matters. You've now got leaders across three major AI agent categories, enterprise automation, customer support, and voice, all certifying against the same standard. That starts to look less like a one-off and more like the beginning of a real industry trend. If you're an operator, your day is a non-stop stream of decisions. And most of them require you to look at the data. You don't need another dashboard. You need answers you can trust fast. But the bottleneck is always the same. The data isn't ready. It's scattered. It's messy. Definitions aren't clear. You're waiting on your data team or waiting on domain experts for clarification and confirmation. That's the bottleneck today's sponsor, PromptQL, is built to break. PromptQL is a trusted AI analyst for high-frequency decision-making. It connects across warehouses, databases, SaaS, and internal APIs. No massive data prep or centralization required. It's built for multiplayer input. Teammates can jump into a thread, correct assumptions and nuance, flag edge cases. PromptQL turns everyday conversations into a shared context. And if something is ambiguous, it doesn't guess. It escalates to the right expert, captures the correct logic, and gets it right next time. That's how it delivers trust and accuracy. Over time, PromptQL specializes to your business, like that veteran employee who just knows things. From simple what-is questions to complex what-if scenarios, you can model impact and stress test decisions before you commit. all through a simple natural language prompt. PromptQL, the trusted AI analyst for teams with shared context and messy data. Welcome back to the AI Daily Brief. Over the last couple of days, we have a bunch of stories which on the face of them are unrelated. It's different companies announcing new products or updates to their old products, all trying to jockey for position in the ever-changing AI landscape. And yet, when you look at all the announcements, there is clearly a convergence happening. The products are starting to mirror one another. We've discussed a version of this trend as the clawfication of AI, but it feels like there's something even more going on. Here's how Buco Capital summed it up. OpenAI is building a super app, bro. It can do everything. And Lovable can do general tasks now. It also does everything. Airtable pivoted. You can vibe code there now. I send all my agents to my Mac Mini to fight to the death and I'll use the strongest one. Bro, AGI is here. So let's talk about what OpenAI's plans to launch a desktop super app, Google's release of their new Vibe Coding experience in Google AI Studio, Lovable's announcement of lovable general tasks, and Claude Code's announcement that you can use it from Telegram all have to do with one another. The temptation I think is for people to view these companies and maybe the AI product industry more broadly as failing throwing everything against the wall and releasing kitchen sink products that don really make any sense I think though what we actually seeing is a recognition that the capability to code does not just unlock new approaches to software engineering and vibe coding, but basically everything else in knowledge work. But let's go back and start with what was announced from Google AI Studio. Google AI Studio themselves tweeted, So a couple things going on here. First of all, Google is integrating anti-gravity directly into Google AI Studio rather than these things being totally separate experiences. Along with that, they are trying to build a more end-to-end experience where you can actually get all the way to applications that can be deployed, as they put it, going from prototypes to production apps. So a lot of the parts of the announcement are just the boring guts required for that sort of move. Integrated databases and authentication, access to modern web tools like Framer Motion, and connections to external services like databases and payment processors. And yet there are also some very googly parts of this announcement. One of the things that we've been tracking, especially as OpenAI and Anthropic go tit for tat with coding capabilities around Codex and Cloud Code, is that while Google certainly hasn't withdrawn from the AI coding fight, this announcement is proof point of that, they also are clearly trying to compete in areas where they are just in a class of their own, specifically around everything having to do with multimodal. Anything that benefits from having access to the entire corpus of YouTube, for example. We see that in things like the Genie 3 model, and we even see it in the specific ways that they're pushing this new vibe coding experience in Google AI Studio. Specifically around this idea of pushing real-time multiplayer games. This is the first use case that they highlight in their announcement post. And I don't think that that's because they think that there are so many people out there right now who want to build massive multiplayer first-player laser tag games. I think they're trying to show off a capability set that they believe is very different. I started playing around with this a little bit, prototyping a game where you take a design from Leonardo da Vinci's notebooks and can actually interact with it in 3D space, trying to turn it into a working machine, almost as a sort of 3D exploratory sandbox type of Myst game. Now when the first iterations of this game experience weren't as visually appealing as I wanted, I fired up a different new Google tool that had been updated just the day before. That tool is their updated creative canvas called Stitch. On Wednesday, Google Labs tweeted, Meet the new Stitch, your vibe design partner. Now the upgrades that they promised as part of this new version included an AI native canvas, a smarter design agent, native voice integration so you can design by talking, instant prototypes, and transportable design systems. It's really a mass expansion in some ways of what people think of as design. And of course, what's going on behind the scenes is that Google is leveraging these new models' capabilities to code to make a better design experience. A couple days later, they dropped a set of new starter ideas that show how blurry a lot of these knowledge work tasks are getting. Their starter idea number one was to take a messy document and turn it into a fully styled portfolio. And what's clear is that Google has ambition to be integrating and expanding these experiences in very short order. Logan Kilpatrick again writes, And that, of course, is what I think is the broader point in all of these announcements. So what's the next one? The next one is Lovable for general tasks. Lovable CEO Antoine Ossica writes, Lovable has always been for building apps. Today, it also becomes your data scientist, your business analyst, your deck builder, and your marketing assistant. This is a big step towards what Lovable is becoming, a general purpose co-founder that can do anything. Some of the examples they show to show off the new tools, including dropping in a CSV file of health industry data to find a startup idea, taking an application that you've built in Lovable, and then creating marketing assets to help launch it, or creating a pitch deck for that app. Now what's interesting is that this is actually quite similar to what Replit announced with Replit Agent 4 a couple weeks ago. In his announcement tweet, Replit CEO Amjad Massad wrote, software isn't merely technical work anymore, it's creative. Introducing Replit Agent 4, design on an infinite canvas, work with your team, run parallel agents, and ship working apps, sites, slides, and more. So let me show you an example of how these things are all blending. What you're looking at right now, or hearing me describe if you're just listening, is effectively a slides as a webpage view of our February AIDB usage pulse survey. Even though the information is still conveyed in slides, you can interact with it like it's a website. Basically, I built the website version and the downloadable slides version at the same time using Replitz Agent 4. And it turns out that this pattern of the blurring of information output is not something brand new just being explored by these companies for the first time. For example, when you're working in Gamma, when you start something new, you have the option to create a document, a presentation, a mobile experience, or a web page, or you can do it all at the same time. When you are using GenSpark or Manus to build slides, what's happening behind the scenes is that their general agent is using code to deliver against anything that you're actually looking for as an output. In other words, the GenSpark general agent is a coding agent with the coding part abstracted and the output format placed front and center. Which is why I think people are a little off with one of the common responses that I've seen to Lovable's announcement that this is a move of some type of desperation. Adam Barta writes, First sign that Lovable is dead. Pivoting to General Assistant is the most investor-pleasing move you could do. Their app-building business is obviously going nowhere and investor money is drying up. Why should anyone use Lovable instead of the already established ecosystems? Now, for what it's worth, just a week ago, Lovable reported that its ARR jumped from 300 to 400 million in a single month. So I'm not sure that it's fair to say that its app coding business is going nowhere, but Adam's hardly alone in this sentiment. Tyler Angert writes, this is the founder equivalent of becoming a paperclip maximizer. Increase shareholder value, they said. We must increase our TAM to 8 billion. Therefore, we will literally make our core product a kitchen sink for general purpose work. Why? Just make separate products if you are so inclined. What a completely dilutive move Going as horizontal as possible with no opinion Hardik Pandaya writes complete strategic dilution may not go well It a huge reach to go from building apps to doing anything a business needs Now of course not everyone agrees Prajwal Tomar writes people say lovable is spreading too thin by going beyond code, but think about it. You need to build the MVP, analyze user data, pitch investors, and run marketing. It just became the tool that does all of that in one place. No more jumping between five different AI tools. This saves so much time. And while that's a totally reasonable argument about the product value here, I think Peter Yang has the right of it. When he writes, and in this case this was after the Replit Agent 4 launch, code is the foundation of all knowledge work. If an agent can write code, it can also generate apps, presentations, animations, and more. Indeed he resurfaced with that same sentiment around the lovable announcement writing, code is the foundation of all knowledge work. Another proof point right here. Now, this is, of course, something that we've talked about on this show before. In January, I did an episode called Code AGI is Functional AGI about why the advances in coding capability mattered, not just because of the way that it would impact software engineering or even Vibe coding tools, but because of the other capabilities that it unlocked. And providing a little evidence that even thinking about Vibe coding as its own category might be increasingly reductive, one interesting finding from our AI Usage Pulse survey for February, which admittedly is at the very vanguard of users given that it's all of you guys answering, who as listeners to a daily AI show are not going to represent the average human being, let's just put it that way, still 71.3% of respondents were vibe coding in February. 62% had some use case that went beyond just assistant into the realm of automated or agentic AI. And while we saw coding use cases continue to be the most common and highest reported value use cases, we also saw a real diversification from coding into other strategic knowledge work areas like data analysis and strategic planning. For some, what's happening is just completely inevitable. Wabi creator Eugenia Kudia writes, Ben Vinegar puts it more poetically, Which brings us to OpenAI. On Thursday night, The Wall Street Journal released an exclusive report about OpenAI's plans to launch a desktop super app that would combine ChatGPT, Codex, and their browser into a single experience. The WSJ points out that the strategy marks a shift from OpenAI's previous approach to launching lots of standalone products that all had to stand on their own two feet. Now, this, of course, gets back to those comments from CEO of Applications, Fiji Simo, where she told the company that they were going to stop focusing on side quests and spreading their efforts across too many different areas. Peter Yang again writes, I think OpenAI's strategy is pretty clear. One, more people have ChatGPT installed than any other AI product. Two, make ChatGPT great for coding and knowledge work. Three, make it a personal assistant like OpenClaw that knows you and can do whatever you want. They just need to get to two and three faster before people switch to Claude or Gemini for the same use cases. Swix, aka Sean Wang from Latent Space, pointed out, meanwhile, that a very long time ago, he had written a blog post with the line, Attempts at building super apps have repeatedly failed outside China, but it's clear that both ChatGPT and Claude Cowork are well on their way to being AI super apps, except instead of having every app having their own app, they make themselves legible to the AI overlords with MCP, UI, and skills, and OpenClaw markdown files. Speaking of OpenClaw, one of the other things that we've been watching is the way that Anthropic has been slowly going one by one through the features of OpenClaw that people like and adding them into the core Claude Code or Claude Cowork experience. The most recent announcement on that front comes from Tariq from CloudCode, who writes, We just released CloudCode channels, which allows you to control your CloudCode session through select MCPs, starting with Telegram and Discord. Basically, you can now message CloudCode directly from your phone, which was of course a huge draw for the OpenClaw experience. Now, Gagan Saludja thinks that this shows OpenAI and Anthropic heading in slightly different directions. He writes, OpenAI merging ChatGPT, Codex, and Atlas into one super app, while Anthropic ships features like channels, persistent memory, and 10k skills in the same month. Two very different strategies playing out in real time. One is consolidating everything under one roof, the other is making the core tool so extensible that the ecosystem builds itself around it. And while he may be right that there is a slight difference in strategy, I think that might have to do more with the starting point of where they are, in other words OpenAI having to deal with product sprawl, than actually being a different strategy. It feels a little bit like both ends are working towards the middle here of a very similar type of experience. Indeed, certainly Fiji Simo herself seems to suggest that this is more about having a Codex Plus experience than it is about having Codex sit alongside a bunch of other experiences. She writes, companies go through phases of exploration and phases of refocus, both are critical. But when new bets start to work, like we're seeing now with Codex, it's very important to double down on them and avoid distractions. Really glad we're seizing this moment. Put differently, it may not be that Open AI is trying to create a super app, it's that they believe that inherently Codex is their super app and they're organizing everything around it. Now, even if I'm right, and this convergence does not show flailing and a lack of product vision, but instead a natural path from coding capabilities to broader knowledge work capabilities, that still doesn't mean that the everything app approach will actually work from a product standpoint. Amwill writes, on one hand, I will be happy to have GPT Pro and Codex, but on the other, I've really come to appreciate all the focus and attention they've placed on making a purely software engineering-focused product. And I think it is worth noting that the other thing that's going on here is just the first large-scale startup competitions in an era where there are officially no moats. Ed Sim writes, When shipping new features costs near zero, every company becomes every company. And when switching costs are also near zero, who wins? The next few months are going to be interesting. I think it's more than the next few months. I think that we are in a totally different type of company-building paradigm that we have barely wrapped our heads around. On the one hand, there are no barriers to entry. People can build and spin things up faster than ever before. Non-technical founders can build the early versions of their products. And yet on the other hand, basically all the traditional moats have fallen. No barriers to entry but also no moats is a very strange and kind of viciously competitive environment that makes continual pivots feel like the only operational strategy. In AI land, nothing is going to sit still for long. For now, if nothing else, we have a lot of fun new toys to play around with, and for that alone, I am grateful and excited. For now though, that is going to do it for today's AI Daily Brief. Appreciate you listening or watching as always, and until next time, peace!
