Forking Paths: Anthropic's IPO Choices  🔗 **Source:** [Forking Paths: Anthropic's IPO Choices](https://share.snipd.com/episode/1baa6b2d-cf89-497c-8360-001738ce6625) ## Highlights - **Consolidate AI Tools To Cut Costs And Complexity** - If you use many AI tools, consolidate them to save cost and streamline workflows with platforms that bundle multiple models and automation builders. - Jaeden recommends his startup AI Box which offers 70+ models and a no-code automation builder for $8.99/month. Transcript: Jaeden Schafer Before we do, if you're someone who works with AI regularly, and you're bouncing between different tools, you should absolutely check out AI Box, my own startup at AI Box.ai. We give you over 70 models in one place. And the part that I think is really underrated is the automation builder. You just describe what you want in plain English, no code, no workflow, no diagrams, and it will build the tool for you. It's $8.99 a month to get access to all 80 of the top models. That's image, audio, video, text. You also get our no-code AI app builder. This should save you a ton of money compared to paying for all of these different tools. Separately, I've been using it and it's becoming one of the things I cannot live without. There's a link in the description to AI box.ai. ([Time 0:00:50](https://share.snipd.com/snip/1ae24771-908d-4c6d-b3a3-e119e05e0132)) - **Apple Will Ship Simple AI Smart Glasses** - Apple is testing four non-AR smart glasses designs that focus on media, calls, photos, and built-in AI rather than mixed reality displays. - Jaeden Schaefer cites Bloomberg reporting that Apple aims to sell them by 2027 with oval/circular frames and colors like black, blue, and light brown. Transcript: Jaeden Schafer The first story I want to talk about today is Apple testing new diagrams for smart glasses. You know, I'm a huge sucker for smart glasses. I'm not even a huge meta stand throughout my life, I would say, but I think that their smart glasses are phenomenal. I think a lot of other players in the industry are noticing that according to Bloomberg's Mark Gurman, Apple is actively testing for different frame designs for smart glasses that They're going to they're like their goal is to go sell these in 2027. I mean, obviously, this is a good play. They have the technology because they have the Apple VR headset, which I honestly forget it exists until I have to go to the Apple store to get something done with my laptop recently, I had to do that. And I saw them on display and people were trying them on. I just feel like those were a really big flop as far as a product goes. Meta might have kind of baited them into the industry, but the glasses are not a flop. I know a ton of people that will actively wear Meta Ray-Bans around with AI and cameras embedded in them. And so I think Apple knows that this is a good play. According to Tim Cook, there are two oval or circular options. There's a bunch of different sizes. They're also looking at black, blue, light brown. And I think what's interesting is these aren't going to have displays. There's no AR overlays. There's no kind of like mixed reality, even though they have the VR headset. They're going to basically let you take photos and videos. You can answer calls. You can play music. And you can interact with, of course, their new upgraded AI series. So functionally, I think it's a lot ([Time 0:01:28](https://share.snipd.com/snip/281312a0-6eab-4daf-9a2c-823273863067)) - **AI Agents Are Powering Vercel's Growth** - Vercel's CEO says AI agents are creating a substantial share of apps and the company is readying for IPO as ARR jumped from ~$100M to a $340M run rate in months. - Jaeden Schaefer notes ~30% of apps on Vercel come from AI agents and his personal projects were spun up via Claude integrations. Transcript: Jaeden Schafer Something else that I thought was very fascinating right now, Vercel's CEO is Guillaume Rauch, which if you don't follow him on X, by the way, he's a legend and drops some really great Insights into what's going on with AI. So I highly recommend giving him a follow. He recently said at the company's annual recurring revenue has gone from about a hundred million dollars at the start of 2024 to a run rate of $340 million by the end of February this year. And when he was asked about an IPO, he basically said that Vercel is quote, very much a working public company and that it's, quote, ready and getting more ready every day. Personally, I don't think any of this is surprising. In the last two weeks, I think I've spun up five different projects where I've moved them off of other hosting providers and put them on Vercel because it integrates so well with Claude Code. And if you ask Claude Code, like, where should I host my, you know, Claude Code app? It's going to say, well, why don't you try Netlify or Vercel? And Vercel is, you know, a slicker looking company than Netlify. So it's a no brainer, right? Like, what features should they have? I don't know. But Claude Code recommended it. So I set up five of my websites on Vercel. In all seriousness, though, I'm on their free tier, I'll probably get bumped up to their paid tier here soon. But I put so many projects on there for cells in the absolute legend. And because it has a API integrates with cloud code very easily. I don't go into verse cell and set up anything. I don't know what the inside of it looks like. I don't know how to, you know, point my name servers there. I don't have to do any of that. I just tell cloud cowork, hey, you know, I'm building an app, go to my domain registrar, point it at verse cell, get everything set up for me, and it does it all. So Vercel is a huge winner in this AI race. And it's kind of interesting, because obviously, Anthropic is coming out of the gate and just killing a bunch of software industries. There's also a bunch of, I feel like their infrastructure, right, like servers and databases, right, you have like Supabase. There's a lot of these companies that I think integrated with AI are just phenomenal companies. And I mean, look, 30% of the apps running on Vercel's platform right now came from AI agents, not humans writing code. 100% of the apps that I have on Vercel are coming from agents, not me, because there's no way I want to do that. So I think this is definitely the direction ([Time 0:03:12](https://share.snipd.com/snip/bd94039a-09c6-4d5a-9ab4-91854e372634)) - **OpenClaw Creator Suspended After Pricing Change** - Anthropic temporarily suspended OpenClaw creator Peter Steinberger for "suspicious activity" after policy changes that separated Claude subscriptions from third-party tool use. - Jaeden describes OpenClaw's virality, the pricing shift, and his personal preference for Claude Cowork instead. Transcript: Jaeden Schafer The next thing I want to talk about, speaking of Anthropic, is their temporary ban on the creator of OpenClaw. So Peter Steinberger, he created OpenClaw, basically, you know, open source AI coding tool. It went super viral. He recently posted on X and he said that Anthropic had suspended his account for, quote, suspicious activity. Basically, this happened shortly after Anthropic changed their pricing so that Claude subscriptions no longer covered usage through third party tools like OpenClaw. And of course, he created OpenClaw was kind of hired or acquired by OpenAI. Users now have to pay separately through the API. So they can't just use the Claude Mac subscription, which to be fair is a subsidized subscription. That's what I have. I don't use OpenClaw because I'm using Claude Cowork and Claude Code kind of combined to get a lot of my projects done. CloudCowork is amazing. OpenClaw, I'm sure, is amazing too, but CloudCowork is amazing because it takes control of your computer. It has computer use. You tell it to do anything. It goes and sets up your Vercel server, your websites, your hosting, your databases. It can go edit videos on CapCut. I mean, honestly, it's insane. The problem with it, though, is that if you're on the $200 a month tier with Claude Cowork, you're really using thousands of dollars of credits, and they're just kind of subsidizing It right now for early users. And they don't want to subsidize people that are using OpenClaw. So honestly, I know they actually get a lot of flack for this, because it's like, oh, they're fighting against open, you know, source. And by saying that everyone has to use like their API to run it on OpenClop. I mean, it's subsidized. So at the end of the day, it's it's what's kind of more profitable for their company. So I don't know, I'm not too too ([Time 0:05:23](https://share.snipd.com/snip/76361f9c-5b88-4b61-b127-227f49485b8b)) - **Government Pushes Banks To Try Anthropic's Mythos** - US Treasury and Federal Reserve urged major banks to test Anthropic's Mythos model for vulnerability detection despite legal and supply-chain tensions with the company. - Jaeden highlights banks like JPMorgan, Goldman, Citi, BofA, and Morgan Stanley being encouraged to test Mythos. Transcript: Jaeden Schafer The next thing I want to talk about is that there is the Trump officials are apparently encouraging banks to test Anthropics mythos model. This is a model that went super viral because it's like, you know, it can go and hack like all of this different software. We've recently had a lot of reports come out that say, you know, Anthropics saying it was too dangerous to release was probably a little bit overblown and probably like a big publicity Stunt. But apparently, Treasury Secretary Scott Bessett and Federal Reserve Chair Jerome Powell brought banks and executives in for a meeting this week, and they encouraged them to use This new model to detect security vulnerabilities. JPMorgan Chase was already listed as one of the original partners from Anthropic when they kind of announced this. But now Goldman Sachs, Citigroup, Bank of America, Morgan Stanley, all of them are going to be testing it out too. What I think is interesting is that Anthropic is currently in a court battle with the Trump administration. And the Department of Defense has designated Anthropic as a supply chain risk after some of their negotiations broke down over limiting military use of the model. Anthropic put some terms in saying, hey, look, the military can't do these things with it. The military wasn't happy about that. So they're having this kind of legal battle. And at the same time, when it comes to patching security vulnerabilities, like, well, we'll put our little legal battle to the side, make sure to use their latest and greatest model For the banking system, which honestly, I mean, I'm as a user of the banking system, I'm happy that this is happening. I think it's not just a US thing, though. The Financial Times reported that the UK financial regulators are also discussing the risks posed by Mythos. They're talking about it from a different angle, though. They're concerned about what it means that the model might be good at finding vulnerabilities that Anthropic won't release. And that's why Anthropic isn't releasing it publicly. ([Time 0:07:24](https://share.snipd.com/snip/929823fe-bb0c-4108-ba8e-99c725c3b9c6)) - **AI Leadership Is Becoming A Personal Flashpoint** - The New Yorker profile of Sam Altman and a Molotov cocktail attack on his home illustrate escalating personal and political risk for AI leaders. - Jaeden notes Altman's public reflection, admission of mistakes, and worry that narratives can incite real-world violence. Transcript: Jaeden Schafer Okay, the wildest story I think that happened in the last few hours is that Sam Altman has posted publicly, he responded to the New Yorker investigation, there was an attack on his home, Which I think is absolutely wild, definitely very unsettling. Early on Friday morning, this kind of someone came in through Molotov cocktail at Sam Altman's home in San Francisco. Nobody was hurt. The suspect was later arrested at opening headquarters where he was threatening to burn the building down. So obviously mentally unstable person. This came just a couple days after the New Yorker published a really long investigation by Ronan Farrow and Andrew Martins, two very serious journalists. Farrow won a Pulitzer for his Harvey Weinstein reporting. And they basically interviewed more than 100 people with knowledge of Sam Altman's business contacts. And the picture that they painted, I mean, I saw a bunch of clips of them kind of talking about it on like podcasts and social media. And it was definitely very unflattering. Most sources describe Sam Altman as having what they called, quote, a relentless will to power that set him apart from other like tech CEOs. One anonymous board member from OpenAI was actually quoted, and they said that he combines, quote, a strong desire to please people to be liked in any given interaction with a sociopathic Lack of concern for the consequences that might come from deceiving someone. So, I mean, obviously this is very heavy. I think it's pretty consistent with reporting from other journalists who've also profiled Sam Altman over the years. There's definitely this kind of pattern of questions about whether the public persona matches how he's behaving in private behind closed doors with his board. I mean, a lot of the stuff that I saw specifically was him kind of going to the board and like just flat out lying about things. Sam Altman published a blog post Friday night responding to both the attack, you know, someone attacking his home and the article. He acknowledged that he had made mistakes, particularly around being quote, conflict averse. I mean, you know, he's basically saying sometimes he will be conflict averse and lie to avoid conflict. I think he said that that's not great. He specifically referenced the 2023 board drama, which is, you know, when he got removed and then reinstated the CEO, he called it something that he's not proud of handling of how he Handled it. He said he handled it badly. Said that he's quote a flawed person in the center of an exceptionally complex situation trying to get a little better each year it's not getting better every day but he's getting better Every year he also drew a really direct line between the article and the attack in his like blog he said that someone had warned him that this article was going to come out and that you know They're like look it's gonna be super dangerous for you because a lot of people are mad about AI. And now there's this. So you're going to be in danger. And apparently he was like no way. Like I'm going to be fine. And then he was like he said quote. He had underestimated the power of words and narratives. In his post he also talked about what he called the ring of power dynamics. In the AI industry. It basically was kind of idea that the prospect of somebody controlling AGI. Makes people do really extreme things. And then he kind of made the point that the solution isn't for any one person or company to hold that power, but to quote, orient towards sharing the technology with people broadly, which Obviously is a very interesting statement from the CEO of a company that was previously an open source nonprofit and has since become a closed source for profit, very valuable, you Know, one of the most valuable private companies in the entire world, the most valuable private company in the entire world. So I think there's a few things that happened here simultaneously. First, obviously, the attack itself is super disturbing. I mean, whatever your opinion of Sam Altman or OpenAI, obviously, no one shoved their home firebombed. That's super messed up. And it reminds me of like, it's probably even a more extreme level. But I remember when Elon was getting involved in politics, and arsons were going and throwing Molotov cocktails inside of like Tesla, you know, stores and burning them down and people Are there. And it's like, obviously, this is very messed up. It's vigilante. It's, you know, it's anarchist, it's illegal and dangerous and like terrible. ([Time 0:09:10](https://share.snipd.com/snip/e23e0952-bc2e-4520-9292-e3a3db146b67))