…[Gabriel Petersson] is a researcher at OpenAI, working on the Sora team — the people building the AI video models that are currently blowing everyone’s minds.
He didn’t get there because he had the right connections or a shiny Ivy League diploma.
He got there because he realized something early on that most of us take decades to figure out: School was just a side quest. Yes you heard right, he took school as a side quest. — Read More
Recent Updates Page 14
How the Internet Dies
For years, “Dead Internet Theory” was framed as something done to us. Foreign bot armies. State-sponsored troll farms. Algorithmic propaganda flooding social media from the outside. A clean external villain you could point at, sanction, and try to take down.
That’s not really what’s happening. The platforms are doing it to themselves. Sometimes through outright bad-faith decisions. More often, through the kind of strategic confusion that looks identical to bad-faith from the outside.
In the last two years, four of the most human-feeling corners of the web (Pinterest, Reddit, Steam, Discord) have each made a series of decisions that are gutting the very thing that made them work. Some of those decisions are pure villainy. A lot of them, honestly, are just woefully bad judgment dressed up as strategy. The result is the same either way.
This isn’t really a “the internet is dying” piece. I genuinely don’t know what the internet looks like in five years. I’m pretty sure it’ll be very different from what we have today. But the mechanism is now visible enough that it’s worth writing down, because once you see the pattern, you can’t unsee it on whichever platform you open next. — Read More
Demis Hassabis: Agents, AGI & The Next Big Scientific Breakthroug
Behind the Scenes Hardening Firefox with Claude Mythos Preview
Two weeks ago we announced that we had identified and fixed an unprecedented number of latent security bugs in Firefox with the help of Claude Mythos Preview and other AI models. In this post, we’ll go into more detail about how we approached this work, what we found, and advice for other projects on making good use of emerging capabilities to harden themselves against attack.
Just a few months ago, AI-generated security bug reports to open source projects were mostly known for being unwanted slop. Dealing with reports that look plausibly correct but are wrong imposes an asymmetric cost on project maintainers: it’s cheap and easy to prompt an LLM to find a “problem” in code, but slow and expensive to respond to it.
It is difficult to overstate how much this dynamic changed for us over a few short months. This was due to a combination of two main factors. First, the models got a lot more capable. Second, we dramatically improved our techniques for harnessing these models — steering them, scaling them, and stacking them to generate large amounts of signal and filter out the noise. — Read More
New in Claude Managed Agents: dreaming, outcomes, and multiagent orchestration
Today we’re launching dreaming in Claude Managed Agents as a research preview. Dreaming extends memory by reviewing past sessions to find patterns and help agents self-improve. We’re also making outcomes, multiagent orchestration, and webhooks available to developers building with Managed Agents. Together, these updates make agents more capable at handling complex tasks with minimal steering. — Read More
OpenAI Flips the Script
If you’re looking for evidence of AI’s unrelenting pace, here it is: In January, Dan wrote that whoever wins vibe coding wins how you work on your computer—and that OpenAI had some serious catching up to do.
Three months and the release of OpenAI’s latest model later, Codex is there, and in a new episode of AI & I, Dan and Austin get into why they do much of their knowledge work in Codex now. They cite the power of GPT-5.5, paired with a desktop app that is faster and more powerful than Claude Desktop or Cowork. — Read More
How AI agent memory works.
A language model on its own is stateless. You feed it a prompt, you get back a continuation, and the moment the response is finished the model forgets you ever existed. There is no “previous conversation” living inside the weights.
An agent, by contrast, is the orchestration around the model: a loop that decides what context to pass in next. Memory is the part of that loop that carries information forward. Everything in this essay is a different answer to the same question, what should we put in the prompt this time? — Read More
Beyond the hype: The enterprise AI architecture we actually need
he real future of enterprise AI is a structured architecture of private models and agent orchestration that works for teams without a complex training program.
My last few years working as a chief digital officer have been, in large part, a sustained exercise in separating what enterprise AI can actually do from what we as a world insist it is about to do. That distinction is not academic. It is the difference between a transformation program that delivers and one that produces a glossy internal report and a quietly shelved proof of concept.
Enterprise experimentation with generative AI has accelerated sharply over the past two years. The Stanford AI Index reports that more than half of organizations globally are now actively exploring or piloting AI-driven workflows — a signal that the conversation has moved from curiosity to operational pressure for many CIOs.
What follows is not a vendor blueprint or prediction. It is a working architectural sketch shaped by real enterprise constraints — the kind that has to survive contact with a real organization’s data governance function, its compliance team and its late-night incident queue. — Read More
Democratizing Machine Learning at Netflix: Building the Model Lifecycle Graph
As Netflix has grown, machine learning continues to support our ability to deliver value to members and drive excellence across multiple areas of our business. When Netflix began investing in machine learning over a decade ago, it was primarily focused on a single domain: personalization. Scala was the industry standard, our ML teams were relatively small, and optimizing member engagement was our primary use case. Fast forward to today, and machine learning has become the backbone of Netflix’s business transformation. We now apply ML across various business domains.
… Each domain operates with a different tech stack, different business metrics, and a distinct organizational structure. While this diversity is a testament to how machine learning has evolved to drive value across many verticals at Netflix, this growth introduces a new challenge: enabling cross-pollination of models and data across domains. — Read More
Rewiring the C-suite: The fast track to 2030
2026 is the year CEOs must rewire the C-suite—redesigning how decisions are made, how authority is distributed, and how AI reshapes influence—while preserving the decisiveness and clarity enterprises need to move fast. Getting there takes proactive leadership. CEOs will need to work with their C-suite leaders to build execution mechanisms, incentives, and operating models all focused on driving these outcomes.
Our research shows that CEOs who have the greatest success with AI are actively rethinking cross-functional collaboration and embedding AI across end-to-end workflows. They’re building organizations designed to thrive in uncertainty, where productive debate sharpens strategy and smart risk-taking is rewarded.
The 2026 CEO Study’s data, gathered in partnership with Oxford Economics, builds on our study, The enterprise in 2030, which identifies five predictions for the future of the organization. This study’s analysis, informed by our 2030 predictions, reveals five plays that CEOs must make to lead in an AI-first landscape. — Read More
Rick's Cafe AI 