AI and AGI

“The bottom layer is silicone, TPU, plus these resold NVIDIAs. And then they don't really care which chip wins. I mean, obviously, they kind of want their own to win, but they're going to make money either way if you buy NVIDIA from them. The middle layer is being the compute host for Frontier Labs. Anthropic already runs on TPUs, and now Thinking Machine Labs is going to be running on Google Cloud. The top layer is the agent layer in the browser and the workspace. Google is the only player that's hitting all three credibly right now.”

“Google announced two new TPUs. They have the TPU 8T for training, and the TPU 8I for inference. The split alone is, I think, what's really interesting to me. Inference is the dominant cost of running AI and production right now, and having these kind of dedicated inference, silicon matters economically. It makes a big impact to the bottom line.”

“They have a new feature, which is called Auto Browse, Gemini powers it, and it's basically running inside of your Chrome for workspace, and it reads context across your open tabs, and it automates actual workplace tasks. So it can enter your CRM data, it can compare vendor quotes, it can summarize candidate portfolios, it can write up competitor research. If I'm being honest, it sounds like an interesting tool.”

“Google Cloud has just signed a multi-billion-dollar deal with Miriam Moratti's Thinking Machine Labs. This is a single-digit billions. It gives Thinking Machine Labs access to NVIDIA's GB300 system on Google Cloud, plus training and deployment services for their first product, which is called Tinker, which is a tool for building custom frontier models.”

“OpenAI and Infosys just announced a deal to push ChatGPT and specifically Codex. Their coding tool is kind of a competitor to Cloud Code into Infosys' enterprise client base. This is across more than 60 countries. They didn't really disclose the terms or how much money is changing hands in this.”

“There is a report from Bloomberg yesterday that an unauthorized group got access to Mythos, which is Anthropic's new exclusive enterprise cyber security AI tool that's gonna take over the world, right? They gave it to a handful of enterprises because it's so dangerous and good at getting, you know, finding security vulnerabilities. I think this is probably not good news because they, you know, they made a lot of hype about how dangerous it was, and they only gave it to special people, and if it really is getting leaked, that's not great for them.”

“NeoCognition is trying to build agents that self specialize the same way instead of kind of the current model where you hand, you know, craft a custom agent for every vertical. I've built enough custom agent workflows to know that kind of this per vertical approach doesn't really scale. You run out of engineers before you run out of use cases.”

“Apple CEO Tim Cook is officially stepping down on September 1st, and John Tarnas is gonna become the CEO. This is really interesting. There's a lot of AI things at play for Apple, and they really haven't delivered on their AI promise.”

“The attacker actually broke into Context AI, which is an AI tool that Vercel, one of their employees was using. And from there, they got into that employee's Google workspace and from there into Vercel's environment. So there's kind of like this multi-step approach that they took in order to get in.”

“As part of the settlement, Clarify has to delete the 3 million photos and any models trained on them. I think that this is something that is absolutely insane. For years, AI companies have been operating kind of in this gray area that, you know, they're like, well, if it was on the internet, we can use it. And the FTC just said, no, you actually can't.”

“He said that AI is now generating more than 65% of new code at Snap and the company expects to cut over $500 million in annualized expenses by the back half of this year. So they are cutting in a massive way. The stock jumped almost 8% on the news, and I think that tells you exactly how Wall Street is feeling about this tradeoff.”

“Previously, they had Image 1.5. Basically, it's really good at doing something that the last model, 1.5, was pretty bad at, which is rendering text. If you ask it for a poster, you know, it would always get this kind of like elfish looking, I don't know, the letters weren't that great. Anyways, 2 in their words is a step change.”

“So that trend is happening, which is we spend as much time now thinking about the agent interface to our tool as we do the human interface. And the reason we're doing that is because our hypothesis would be that if you have a hundred or a thousand times more agents than people, then your software has to be built for agents. And then what is the way that those agents are going to interact with your system? It's going to be through an API or a CLI or MCP or whatever.”

“The question for every software company is no longer whether to support agents, but what it means when agents outnumber employees a thousand to one. I speak with Aaron Levie, CEO at Box, alongside a16z board partner, Steve Sinofsky, and a16z general partner, Martin Casado. Agents do not want simpler systems, they want better ones. They choose backends based on durability, cost parameters and reliability, not interface polish.”

“Well, then you're basically just developing the next abstraction layer for how people interact. And developing an abstraction layer has historically at each level of the abstraction layer been a highly skilled, very specific individual within an organization developing that. And then the little parts that they build just become little toolets in the world of people doing particular tasks. And some people are able to stitch them together and some can't.”

“I think it's very tempting to be like, these agents are going to code and do X. But I think we're going the opposite way. So I think actually where we started was, we'd take a piece of SaaS software and we'd add AI. Then that's the new AI enabled. But now what are we actually doing? We're like, okay, the SaaS software is still SaaS software, and the agent uses it as a computer because it's actually very good at that.”

“The diffusion of AI capability is going to take longer than people in Silicon Valley realize. It's just absurd to think you're going to vibe code your way to like SAP. All of that domain knowledge, it's not just represented in some well-orchestrated data layer. The engineering compute budget conversation is going to be the most wild one in the next couple of years.”

“Something that shocked me, absolutely shocked me, is that Anthropic has just passed OpenAI in annualized revenue. The run rate is $30,000,000,000 at Anthropic. That's about $25,000,000,000 over at OpenAI. Anthropic was at roughly $9,000,000,000 in annualized revenue at the 2025. They hit $20,000,000,000 in early March and 30,000,000,000 in early April. Since December till now, we go from 9,000,000,000 to 30,000,000,000. This is the absolute greatest run ever.”

“Worldwide app releases in 2026 are up 60% year over year across both Apple's App Store and Google Play. And on iOS specifically, it's up 80%. I think vibe coding is working. Claude Code, Replit, all of these tools have gotten so good that people with ideas and no engineering background are shipping apps. Right now, I believe it truly is the gold rush, and the App Store is basically benefiting in a huge way.”

“They are discontinuing Sora. We've talked about this a ton, but the app officially closes on April 26. They peaked at around a million users, then it collapsed under 500,000 while they were burning somewhere between a million and $15,000,000 a day in compute depending on which report you believe. Then on basically the same day, the shutdown timeline locked in, OpenAI lost three of their major executives.”

“Cerebras has just filed to go public. They have a deal with OpenAI that's reportedly worth more than $10,000,000,000. Their CEO, Andrew Feldman, went on record saying, 'NVIDIA didn't want to lose the fast inference business at OpenAI, and we took that from them.' If even a slice of their inference workload is moving to Cerebras, I think that is the first real crack in NVIDIA's monopoly at the frontier.”

“AI agents went from about 12% success on real computer tasks a year ago to 66% now. This is agents actually navigating software. They're clicking through forms. They're pulling data. They're finishing multi step jobs in real systems. 66% is not good enough to let something completely loose unsupervised, but it's good enough for giving an agent a really narrow bounded job and checking its work, making sure that it's getting stuff done faster.”

“In the end, the customer and the customer alone determines whether you win or lose. It's a very simple equation. If you keep them coming back, you got a good chance. And if you don't, you lose. And, you know, back then, one of the most interesting parts about that store was knowing your customer and knowing your base.”

“All people should ever want is an opportunity because an opportunity is a gift, and it's something that you should really cherish. And every time you get one, make the most of it. I am so honored that in my life, I've had the shot. And when you have a chance, you can really bring the winner within you out.”

“The cost to replace an enterprise platform in this SaaS pocalypse that people talk about is an extraordinary expense. Let's take that cost, and then let's take the cost associated with the human capital doing that instead of something else because the platform was doing the work for you, and then let's add up the cost of the GPU factory and the tokens.”

“And in this environment we're in today, which is changing and it's dynamic like never before, and it's fast like never before, I tell them it is fast, but it'll never move this slow again. So this is the way it's gonna be. We have to learn to get, like, real inspiration from the challenges that are in front of us.”

“If you take care of your customer, you know who your customer is, and you give them what they want just the way they want it, no matter what business you're in, you've got a really good chance. And I have to tell you, the EQ that comes with dealing with 500 customers a day is worth a lot, and I have a feeling the world could use a little more EQ right now.”

“OpenAI just massively beefed up Codex for desktop control, memory, and in-app browser, and over 100 plugin integrations. Basically, this is them swinging directly at Anthropix Claude Code and Claude Cowork, and I think it matters a lot for where editing agents is going to be going in the future.”

“Factory... just closed $150 million Series A at a $1.5 billion valuation. Morgan Stanley isn't going to let some random developer tool run inside their network unless it's built with their compliance and security posture in mind. I think that's basically the gap that factory is filling.”

“You can describe what you want, a pitch deck, a one-pager, a landing page prototype, and Cloud generates a first draft. It can also read your company's code base and design files to apply a consistent design system across all of your outputs, which is actually, I think, the more interesting piece.”

“AI users have 9.4 times higher code churn than non-AI users, and Pharaoh AI found code churn increased 861% under high AI adoption. The productivity gains from AI coding are real, but they're also a fraction of what the output numbers suggest, right? If you're like, I can write a million lines of code a day now... well, a big chunk of it has to be rewritten or fixed.”

“Pi 0.7 can perform tasks it was never specifically trained on by composing skills it learned in other contexts. In some broader testing, the generalist model actually matched specialized models on jobs like making coffee, folding laundry, and assembling boxes. Researchers said that the generalization ability was really surprising to them.”

“A 20 year old man allegedly threw a Molotov cocktail at the gate of Sam's home. No one was hurt, but according to the criminal complaint against the suspect, this was someone who had a document that identified views opposed to artificial intelligence, also had a list of names and addresses of other AI executives, investors, and board members. This is someone who was very clearly concerned about the existential risk that AI posed in his opinion, and so decided to take matters into his own hands and go try to attack Sam Altman.”

“The state of Maine recently passed a temporary moratorium that would ban data centers larger than 20 megawatts until November 2027. There's a suburb of Milwaukee, Wisconsin, Port Washington, which is gonna be the home of one of these big OpenAI Oracle Stargate data centers. That town recently voted overwhelmingly in favor of restricting the building of future data centers. Basically, you have to get voter approval before you do any of these things. Then there's also similar efforts going on in places like Ohio, Missouri, Indiana, Georgia, North Carolina.”

“There's a new report out from Stanford this week, the 2026 version of their AI index, which sort of catalogs various trends in the AI industry. And, basically, their takeaway was that, in The US, people have very low trust in not only AI, but on the question of whether their own government can regulate AI in a responsible way. The global average on that question was 54 percent of, like, do you trust your own government to responsibly regulate AI? In The US, that is only 31 percent.”

“Words have power too. There was an incendiary article about me a few days ago. Someone said to me yesterday, they thought it was coming at a time of great anxiety about AI and that it made things more dangerous for me. I brushed it aside. Now I'm awake in the middle of the night and pissed. So what do you make of the idea, Kevin, that a reason for the negative sentiment against the AI companies and the industry at large is being driven by investigative journalism?”

“I mean, one thing that I've been thinking about over the past few days is, like, this is happening at a time when unemployment is below 5 percent, and the S and P 500 is near a record high. And so if all of this is starting to happen when things are relatively good, economically speaking, in this country, I think the fear and the expectation among the leaders of these companies is that it will get much worse if and when AI does actually start to cause, like, mass disruptions to the labor market.”

“We moved from a classic ML world to a world where we have LLMs generating responses and now we want to move on to a world where actions need to be taken. And when the problem that we are working on is a complex one with multifaceted aspects associated with it, that's where multi-agentic comes into place. So basically, we have a large complex goal, which we have to break down into specific steps and each step is basically narrowed to a specific agent.”

“Chat Concierge is essentially an auto dealership project or application that was deployed out to our auto dealers to basically bridge that experience between dealers and their customers and make it very seamless. This is an auto buying experience that we wanted to make sure that we deliver the right solutions or cars to the right customer needs. It was a multi-agentic chat experience that was brought to the fore with the human in the loop to car buying customers to get the right match.”

“In this specific scenario, there were a multitude of intents, so there had to be one agent that understands specifically this intent and tries to disambiguate by asking clarifying questions back to the customer. That is that narrow job. And then from there on, we had multiple tools that can get executed in the form of different actions that need to be taken based on the intent that comes in. So we have a planner agent that does this discernment.”

“These platforms come into the fore when you are governing agents in runtime. And that's where the massive huge benefit of platforms comes into the fore. This gives the architects of that specific agentic framework the flexibility of focusing deeply on the design, whereas the platform brings in all of the governance and risk compliance that needs to be bounded to make these agents execute safely in any environment.”

“We have a very, very robust model risk office that we work very, very closely with. We have all of the risk and compliance frameworks embedded within the platform, which appear as policies, as guardrails, as security enforcement, and cyber enforcement across our different layers of the platform that get implemented across different threat boundaries of the agents.”

“We've also got reports today that Anthropic is turning down investors that are begging them to let them invest into Anthropic at an $800 billion valuation, which by the way is double what their last valuation was. So honestly, some of these companies that are on absolute tear right now, and I think people can see where a lot of like, the infrastructure is going to be in a similar place. Some of these companies that are growing super fast are just saying no to investors, and they're just not raising money right now, because maybe they think they can do better in the future.”

“Avid is basically, I think, the backbone of how a lot of professional video and audio productions are functioning today. Film studios, news organizations, post-production houses. There's just a huge chunk of the industry that basically runs on Avid. So this isn't like a niche software company. When Avid is doing something, the professional media world is basically coming along with it. So I think the vision of these AI-assisted production workflows, you have intelligent content search that's happening across these massive media libraries.”

“What Antioch is trying to do right here is let robot builders spin up digital instances of their hardware and then connect them to simulated sensors that actually replicate the data that the robot is going to be experiencing in reality. So right now, they're focusing on sensor and perception systems, which is kind of where I think a lot of complexity lives for autonomous cars and trucks. And they're also going to be looking at drones, agricultural machinery, a lot of construction equipment.”

“I think the fact that they're 7 months old, and they have no products, and they're talking about a $2 billion valuation is absolutely incredible. I also think it is logical, given a lot of things in the environment, right? So I think the conviction in the AI infrastructure layer is really big right now. Investors believe that whoever controls the next gen of AI compute infrastructure is going to capture, I think, just so much value. So I think a lot of VCs, a lot of investors are placing bets before there's a product.”

“AI Box is my own software platform. It's how I solve this. It's one subscription. It's $8.99 a month, and you get access to over 80 AI models in one place. So Gemini, Claude, Grok, ChatGPT, 11 Labs for audio, all of the top image models, VO3 for video, Sora for video, all of the top models in one place. So what I keep coming back to on top of all of that is that we built an automation builder. You describe a tool or a workflow that you want in plain English.”

“Apple is actively testing four different frame designs for smart glasses that they're like their goal is to go sell these in 2027. I think this is Apple basically accepting reality that the Vision Pros did not land the way they hoped, and so they're going to hopefully move into something that has a lot more market appeal and people are a lot more excited about.”

“Vercell's CEO is Guillaume Rauch... he recently said the company's annual recurring revenue has gone from about $100 million at the start of 2024 to a run rate of $340 million by the end of February this year. When he was asked about an IPO, he basically said that Vercell is 'very much a working public company' and that it's 'ready and getting more ready every day.'”

“Anthropic temporarily banned the creator of OpenClaw... basically, this happened shortly after Anthropic changed their pricing so that Claude subscriptions no longer covered usage through third party tools like OpenClaw. 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.”

“Trump officials are apparently encouraging banks to test Anthropics' Mythos model. 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.”

“Early on Friday morning, someone came in through a Molotov cocktail at Sam Altman's home in San Francisco. Nobody was hurt. The suspect was later arrested at OpenAI headquarters where he was threatening to burn the building down. This came just a couple of days after the New Yorker published a really long investigation.”