Garry Tan

The one skill everyone who uses AI needs to master in 2026 Meta-prompting: using an LLM to generate, refine, and improve the very prompts you use to get work done. https://garryslist.org/posts/metaprompti…

Here is my latest iteration of my /plan-exit-review skill for Claude Code. I use this as I exit plan mode and it works super well to shake out all issues,, shake out architecture and code smell issues, perf issues, and finally make sure every part of a PR is tested (I avg 1.3 lines of test to 1 line of real code these days) https://gist.github.com/garrytan/001f907…

Data Center NIMBYs Are Killing $1 Trillion in AI Infrastructure Meanwhile places like Loudoun County that said yes got $111M in school funding and lower property taxes. https://garryslist.org/posts/data-center…

Software engineering accounts for nearly 50% of all AI agent tool calls. Healthcare, legal, finance, and a dozen other verticals are barely touched, each under 5%. That's a hundred AI unicorns waiting to be built. https://garryslist.org/posts/half-the-ai…

The most important question for builders today is not building Web 2.0 faster The most important highest alpha thing to focus on is what can be built given now you have direct access to LATENT SPACE Intelligence is on tap. What can be done now that could never be done before?

Boston has the schools. SF has the companies. 21 of the Forbes AI 50 founders went to school in Boston—then left for San Francisco. If SF doesn’t learn from this cautionary tale, Austin and Miami will. https://garryslist.org/posts/boston-cook…

AI just moves the excuses out of the way. You don’t get to say “I can’t edit” “I don’t know how to code” or “I’m not technical” Intelligence is officially too cheap to meter. Your taste is your evals. Your desire to win is your agency. Time to build. https://garryslist.org/posts/ai-didn-t-k…

Anthropic’s War on Its Own Power Users Charging power users $200/month then banning them for actually using it is not ‘abuse prevention’ — it’s the RIAA playbook for AI, and we all remember how that movie ends. https://garryslist.org/posts/anthropic-a…

Integrating Exa web search with Claude Code is wild. In plan mode say: I want to use it. Brainstorm where the agents use it and how. Exit plan mode. It’s done. Enter API key. Agents can search the web. It’s insane.

tfw claude code jokes about how many pr's you did today

If you are too worried about being hated, it'll be hard to make anything that people love

The flow I feel when vibe coding now is not just as good as it was 10 years ago during my startup. It is 6x better because I can keep 6 concurrent workers going all day, each in a plan-write-test-PR-ship loop Sometimes I time travel several hours at a time. No feeling like it

Personal agent software that you install locally is one of my favorite new metas of 2026 Placing agent power on your own computer empowers every user and I’m so here for that https://ironclaw.sh

It's amazing to see the cutting edge jump to recursive self improvement when the majority of developers in the world are still mostly skeptical of vibe coding for production at all

Welcoming PG and Jessica back to YC Back where it all started in Mountain View

Does a hype cycle have a retention curve that looks like this? Literally never seen a retention cohort graph like this

I spent the day using Claude Code macOS app with git worktrees head to head against @conductor_build and Conductor is still better - it's more responsive, doesn't hide what it's doing, more rock solid. Claude Code worktrees is good, but Conductor is still better.

It's like they invented the Star Trek replicator, but for code, and 1% of people are using it

One of the funniest things about Claude is that is sometimes surprised when the user comes up with a legitimately useful idea

AI Is In Its Gentleman Science Era. AI research is wide open for builders with heretical ideas. The window won’t last forever. https://garryslist.org/posts/ai-is-in-it…