I've been surprised and impressed by how all of the leading agents are essentially while loops with llm + tool calls. As a systems nerd, I've always wished that AI software (even pre-LLM) were modular. I love abstractions :) But the reality is, the bitter lesson of AI systems

Is Gen AI before or after Gen Z asking for a fren

idk what happened but my X feed feels better already. Thank you Nikita Bier.

Extreme takes deliver eyeballs, nuanced takes deliver results. You probably need both.

LLMs are the Reader Mode for the Internet.

The start-up version of shavasana is the feeling after the team finished tackling a complex problem and shipped something real.

One easy AI opportunity Apple is missing is a trusted personal assistant that by default has user’s full secure context - both on OS and app level. Imagine asking your Apple AI about health, banking, payments or personal history without trying to scroll through 10 mobile screens

You can’t beat doing important work with people you like.

Chatting to GPT5 and Opus 4.1 about a very public, large digital market and getting directly opposite takes on most common practices. Either we have ways to go to AGI or we'll be able to choose from 5 flavours of it.

how to build a billion dollar startup

Quidditch is a perfect visualization of chasing AI alpha in SF

It never gets old.

You can’t engineer context without understanding the workflows - a great reason to be bullish on industry-focused AI agents.

Dreamforce 🌉- here we go again! Excited to catch up with everyone and talk AI-Native Commerce cc John Boxall SuperpilotAI

🔑 insight from Ed Sim. Successful scaling cos usually run *all* GTM motions to stack growth. Yet, choosing where to start and how to sequence can vary a lot by industry & buying persona.

AI scaling laws are just as exciting as model innovation - even if AGI isn't fully there, cost of superintelligence will trend towards zero on a long enough time horizon.

“Agents not pages” is the biggest theme of this year’s inaugural ShipAI by Vercel. Tectonic shift for the Web as we know it.

Lots of potential counter trends to the "god model" thesis: - proliferation of models at the app layer, even for the same task (e.g. coding) - consensus view on the growing importance of context engineering. - OpenAI agent framework including relatively constrained environment,

Wake up babe 2026 OKR planning just dropped

The AI gift to young talent out there is removing skill barriers to start a business - brick-and-mortar or tech - given how much can now be automated with AI agents. The entry level jobs are still here - the twist is, you get to be the boss!