Nate's Newsletter

Weights, activations, features, and circuits, explained from the moment you hit enter to the answer you get back.

How to find what your agents are consuming, get more work from the machine you own, and choose the right Mac, NVIDIA system, or spare computer when you need more room.

Agents are becoming residents of the internet. The business question is what should survive when a worker disappears: useful knowledge, mistaken assumptions, or permission to keep acting.

Five documented investigations into real software, and how to use what they reveal to build something of your own.

Watch now (21 mins) | Are we developing a more general intelligence, or getting much better at producing extraordinary abilities that transfer unevenly?

Personal apps and agents are expanding what one person can accomplish. Five skills help you grow into that possibility—and they get stronger when you use them together.

Google wants to learn from the record of our working lives. We should take a closer look ourselves, because understanding what our work accomplishes could change what we attempt next.

It’s a phenomenal time to build with AI and a good place to start is with the person whose life you can make better.

Agents are pulling apart the data, screens, and workflows you’ve been buying as one subscription. Each piece now needs its own decision.

Used together, Dots, Space, and Sol turn the context you've built with AI into a shared workspace. The questions to ask are what can it take on and what does that work cost?

The best AI isn’t always employer approved. These are the habits I’d build with the approved tools you have, starting with one prompt you can run this week.

Anthropic heard the complaints about Opus 5. Opus 5.5 and Sonnet 5.5 are both easier to steer, and both make even complex jobs cheaper to run.

A canceled subscription is the beginning. Meta’s new assistant points toward a much larger change in who does the work of managing your life—and who gets to sell you things along the way.

AI is making some people dramatically more productive. Getting that speed through the rest of the organization is a separate problem, and it isn’t solved by making the fast people teach everyone else.

A conversation with NVIDIA’s Cosmos lead about the next expansion of AI: what world models do, how they connect to LLMs, and the strategic choices that matter as AI learns to act.

How do you turn individual AI ability into a team that learns faster, makes better decisions, and gives people some of their lives back? Last week, I went to OpenAI to ask.

Software is about to get intelligent in very different places.

What I heard at Dreamforce points beyond cheaper agents: different customers, different offers, and a different answer to where a company earns its money.

What I took away from my conversation with Emily Sands, Stripe’s Head of Data & AI.

Everyday AI is about to be good enough. That’s the strongest case anyone has made for Apple in two years, and it measures the wrong thing.