Every time I run a huge token-burning task on Muse or Instinct I think about 2021, when I'd order a banana for subsidized 15-minute delivery. Enjoy it while it lasts.

The first 9 months of this year felt like what I imagine the Homebrew era was like, but for personal assistants. People like me spent tons of time hacking together solutions built on OpenClaw or Hermes. The last month has felt like it must have when the Mac launched. Now everyone just gets all of that out of the box in an easy-to-use interface.

Every VC I’ve asked in SF this week how they're approaching this environment has answered with some version of 🤷‍♂️

There seems to be a weird dynamic right now where founders go out to raise rounds as big as the “best companies” raised, because if they didn't, it would signal they’re not one of the best companies.

Lovable is acquiring Sutro! It's been a fantastic experience working with Tomas since leading their pre-seed round. Excited to see the value of what they built recognized by Lovable and that they will now be an integral part of how things get built going forward!

Having invested in early stage startups for 16 years, one thing I've seen up close is that the industry tends to be cyclical but people tend to act in each cycle like it's the new norm. It's a decent chance this time is different, but we should recognize that would be a first.

happy big birthday to a gentleman and a scholar, a friend and a father, a husband and a great partner, and still an up and coming DJ... @Hadley

This is likely driven by a combination of: 1. Rogue agent swarms (see OpenAI x Hugging Face) 2. Recursive self-improvement outpacing alignment 3. Writing the rules before regulators do

I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we've had at OpenAI in recent weeks. Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We'll have more to share soon.

We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so. Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training. You can read the full post here: darioamodei.com/post/we-must…

I tried outsourcing my fantasy football draft to AI and it did terribly. Sorry, Jensen, not at AGI yet!

My simple philosophy for hiring: hire tenacious, high-slope, high-agency people and give them the space and support to succeed.

First day of school is such a special day for many reasons and the fact my boys are finally going to the same school just made life much simpler!

Welp, my month long mini sabbatical has come to an end. Final week at family resorts in Austria, then Munich with friends before flying home. Fun fact: the manager at one resort told me I used the office pod more than any guest ever. Which is funny, because it was only a handful of time sensitive zooms with founders. Coming back super excited. Too many founders doing amazing shit right now to stay away. And I'll leave you with me drinking a beer I bought in a playground on my last night, cause that's a thing in Munich.

Today we're launching OM2. Your AI re-reads your entire company from scratch every time you ask it something. It’s why more than 50% of your token bill isn't in the answer, but in the search for it. OM2 gives your AI a permanent memory of your company. 🧠

Sabbatical week 4: beaches, boats, beats and buddies. This live set dropped somewhere between Ibiza and Formentera on.soundcloud.com/4yqtVSszZt…

"Multi-agent systems" has officially entered the healthcare buzzword hall of fame, right up there with "value-based care" and "interoperability." Everyone's throwing agents at everything and calling it innovation. But every so often, a company actually shows their work. And that's exactly what Predoc (@predoc_ai) did with a new report on how they built the multi-agent architecture behind their medical records retrieval and curation engine. The full report (link below) is a useful blueprint for any healthcare exec trying to figure out what "building with AI" should actually look like in practice or where AI can have the highest impact. 4 things worth digging into: 1. Bespoke work is the opportunity, not the obstacle. Predoc makes the case that the most valuable automation opportunities aren't the clean, standardized tasks. Those get commoditized fast. It's the messy, facility-specific, exception-riddled workflows that are actually defensible. I think that's right, and it's a useful gut-check for any exec evaluating an AI vendor's ROI or worth buying. 2. The dataset is your moat. Predoc built its system on 300K-400K provider-research tasks, nearly 3 years of transcribed retrieval calls, and millions of reviewed record pages. The foundation models are swappable. That accumulated, structured "tribal knowledge" is not. 3. Start from first principles. Break the workflow down into its simplest parts. Bound each job, structure the handoff, escalate the exception. Predoc lays out how they gave each agent a job (research, voice, indexation, extraction, curation) and a structured output the next agent can act on immediately. When something doesn't fit, the agent escalates to a human, and that resolution gets fed back into the system. 4. The numbers back it up. I was pretty intrigued by some of the results in this piece: A 2-week-plus turnaround compressed to a median of 3 business days. Provider-research time down 70%. First-pass retrieval success up nearly 50%. 94.6% of pages indexed without human intervention. The bigger theme I keep coming back to: this is a case study in systems of intelligence sitting on top of disorganized, disparate systems of record. Predoc's real output isn't "faster fax retrieval." It's a normalized, longitudinal clinical data layer that other applications can actually query. Big thanks to brand partner Predoc for sitting down with me and showing their work on this one. predoc.ai/multi-agent-system…

Amazing @truefoundry 's Open Source Harness- TrueForge is trending #1 on twitter!

Big news from our portfolio company @truefoundry, the leading enterprise AI gateway for LLMs, MCPs and agents, trusted by the world's biggest companies. They just open sourced their agent harness, TrueForge. Any model, any MCP server, same accuracy at up to 50% lower cost.