What Vibe Coding is Turning Into

**Nathaniel Whittemore** (0:00)
Today on the AI Daily Brief, what vibe coding is turning into? Before that in the headlines, agents get their own credit cards. 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, Robots & Pencils, AIUC and Blitzi. To get an ad-free version of the show, go to patreon.com/aidailybrief, or you can subscribe on Apple Podcasts. If you are interested in sponsoring the show, send us a note at sponsors at aidailybrief.ai. And as I mentioned yesterday, the latest fun new little project from the AI Daily Brief, you can find it agentmadness.ai. We obviously have got a lot of builders in this community. Many of you have done AIDB New Year. Now some of you are doing Claw Camp or Enterprise Claw, and I wanted a place to show off what you've built. That is Agent Madness. We're basically taking the NCAA's March Madness, i.e. 64 teams competing, moving down through the bracket to the championships, and we'll do a human voting on AI agent system. So go to agentmadness.ai, share the coolest agent you've built, and next week we will build this bracket and get this thing live. Again, that's agentmadness.ai.
They grow up so fast, don't they? It seems like just months ago we were watching agents fumble around, not really successfully doing the things that we asked them to, and now here they are, all grown up, getting their first credit cards. Yes, indeed, both Ramp and Stripe are introducing virtual payment cards designed specifically for AI agents. The Ramp version allows agents to tap into payment options via API, MCP and CLI. If those initials don't mean anything to you, co-check out my episode yesterday about Google Workspace CLI where I go into all of that. In any case, credit card numbers are not exposed through the new workflow, which should improve security, and users can also set spend limits and have real-time visibility into transactions, just in case your open claw has expensive tastes. Ahmada Sino tweets, My Ramp agent has already spent $5,435 on Yu-Gi-Oh cards. No regrets. The Stripe version is just getting started with a limited beta available to agent builders. Once again, the virtual cards have programmable spending limits, controls around merchant categories and real-time risk scores to make sure your agent spending is under control. Stripe product manager Jeff Weinstein explained the logic behind introducing the cards, commenting that we're about to see a lot more agents doing a lot more for us and that's often going to involve spending money. Now this is one of those obvious in retrospect kind of things. You are already seeing people experimenting with handing a credit card to their OpenClaw. Entrepreneur Supermom in front of the show, Jessie Gennett, has recently been going ham on OpenClaw experiments and at the end of February got her OpenClaw wired up to using her Amazon account. But overall, there is still clearly a need for better ways to both enable and control agent spending. One other interesting little detail about Ramp and Stripe doing this is that they're opting for normal cards rather than promoting some agent-focused payment protocol like X402 or AP2. In short, advances in agent capability mean they are more able to use payment rails that were designed for humans than when those protocols were designed six months ago. They also have the huge structural advantage of plugging into existing merchant infrastructure rather than having to bootstrap a new system. Separately, but also from Ramp, the company published a new edition of their AI index, an economist Eric Herazian is ready to call it. I've seen enough, he writes, Anthropic is the new default for businesses. The new update for March shows that AI adoption is growing quickly for Ramp customers across the board, but Anthropic's growth is in a league of its own. Overall, AI adoption is now at 47.6%, meaning almost half of the businesses on Ramp now have at least one AI subscription on the books. Anthropic is the only company seeing accelerating growth and adoption, now at 24.4% of Ramp customers, compared to 34.4% and falling for OpenAI. The big signal, however, that Karazian is paying attention to is Anthropic's share of the first-time AI spending. Compared head-to-head against OpenAI, Anthropic is now winning 70% of first-time business. The two companies were neck-and-neck at the beginning of the year. Writes Karazian, It's a complete reversal of the trend we observed in 2025, when OpenAI adoption accelerated faster than any other model company. Now keep in mind, this is of course just one source of data from one provider, and there are even limits to this methodology. As Adam from OpenAI points out, this might not be fully inclusive of enterprise customers who purchase via purchase order, but certainly this is part of an overall trend line that we have seen numerous examples of. Speaking of OpenAI, the company is planning a bit of a shift around Sora, planning to integrate the model into ChatGPT. The launch of the standalone Sora app in September did garner a lot of downloads, shooting in fact to the top of the App Store charts. Perception, however, was that usage quickly waned, and last fall Sam Altman told staff that only a small percentage of Sora users were sharing their videos publicly, making it not all that effective as a social network. Now one little bit of a narrative violation, given that many people had written off Sora, is that recent data from Sensor Tower showed that over the last month Sora jumped over 3 million daily active users and was continuously growing. And yet still the information reports that by bringing Sora back into ChatGPT, it signals a strategic shift refocusing on ChatGPT as the core OpenAI experience. The information points out that OpenAI has put more resources into ChatGPT recently in hopes of hitting 1 billion weekly active users. OpenAI's leadership is also cognizant of increased competition from Google's Gemini, which now does offer video generation as part of the core app experience. Over in Elon Land, the unification of the Musk Empire continues as Tesla plans a joint AI initiative with XAI. Musk announced on Wednesday that his companies will work together on a project called MacroHard or Digital Optimus. He said the project was part of Tesla's recent deal to invest in XAI. In a post on X, he wrote, Grok is the master conductor and navigator with deep understanding of the world to direct Digital Optimus, which is processing and actioning the past five seconds of real-time computer screen video and keyboard mouse actions. Grok is like a much more advanced and sophisticated version of turn-by-turn navigation software. You can think of it as Digital Optimus AI being system one, instinctive part of the mind, and Grok being system two, thinking part of the mind. Musk noted the system will run on Tesla's low-cost AI for silicon rather than requiring expensive NVIDIA hardware. Stripping out the sales pitch, it sounds like this will be XAI's computer use play, perhaps joining the rest of the industry in the open qualification of everything. However, Elon of course positioned it as a more grandiose vision, adding, it will be the only real-time smart AI system. In principle, it is capable of emulating the function of entire companies. This is why the program is called MacroHard, a funny reference to Microsoft. No other company can yet do this. According to Business Insider, the MacroHard project has run into multiple roadblocks at XAI since it was first mentioned in August. Sources said that two project leads left the company last month in quick succession, after Musk expressed disappointment in a lack of progress. They also noted that a massive data annotation project attached to MacroHard was paused last month to allow for architectural changes to the underlying model. So despite some negative behind-the-scenes reporting, we'll see if the axiom holds that you should never bet against Elon Musk. Lots of chatter about Ben Affleck's big payday. Netflix could pay up to $600 million for his AI startup Interpositive, which would make it the largest AI media acquisition to date. Now, last week's news, that Ben Affleck had sold his AI startup called Interpositive, was already a big story in Hollywood. It came as a shock to many that Affleck was secretly working on AI for filmmakers, given his comments on AI, where he basically thought that it couldn't create genuine works of cinematic art. But of course, there is both more nuance to what Affleck has always said, and more nuance to what Interpositive does. Affleck was never a throw-the-baby-out-with-the-bathwater kind of AI critic. And in fact, in all of the comments that I've seen of his, it was less about what AI couldn't do, and more about what humans were uniquely capable of. Interpositive's work is similarly nuanced. The company is working on custom-trained model for individual productions. The technology reportedly was designed to allow filmmakers to quickly edit shots by automatically adjusting lighting, reframing the camera and replacing backgrounds. It was not, however, designed to generate entirely new scenes, keeping creative control firmly in the hands of the human filmmaker. Essentially, it's an AI that is fully embraced as a cost-cutting tool for post-production, rather than as a replacement for human filmmaking. Bloomberg reports that the technology is already being used by David Fincher for an upcoming film with Brad Pitt. Now, it was already an interesting acquisition, but the sky-high price tag puts the deal in a very different context. This would now be among the largest deals that Netflix has ever done. Affleck is also joining Netflix as an advisor as part of the deal, so my guess is that this is not the last we've heard of this. As I've said before on this show, I actually think that the entertainment industry has a unique opportunity to carve a path showing that AI doesn't have to be a choice of humans on the one side or robots on the other. Call me naive, but I'm going to choose to be optimistic that this type of deal can be part of a future where we get more amazing entertainment content enabled by AI that isn't just cost-cutting slop. Lastly today, a massive jump in revenue for Lovable. The company has added a staggering 100 million in annualized revenue in a single month. Lovable told Business Insider that ARR jumped by a third in February, rising from 300 million to 400 million. Coupled with Cursor doubling ARR to 2 billion over the past three months, it's clear that we are very much in a rising tide lifts all boats kind of moment. Indeed, Lovable's Chief Revenue Officer Ryan Meadows said, It's a rising tide. We've been super happy with what we're seeing. Lovable also launched their debut brand campaign this week, which features a normal woman going through her day with a song rattling around in her brain. She eventually arrives home, fires up Lovable, and prompts it to build an app for the songs in my head. At no point in the ad anywhere does it mention AI or vibe coding. It is instead a story of the way that these new tools collapse the space between idea and something real. Now, speaking of vibe coding, that is going to be the topic of our main episode, so that is where we will close the headlines.