Jason Cameron

I don't mind if you're using AI when contributing to OSS. In fact, I prefer that you do! HOWEVER Please don't just forward messages between me and your Claude. I have my own Claude I can talk to. I don't want to talk to your Claude. Thank you for your attention to this matter.

The only reason I won't stop reviewing the agent's code is that I've reviewed the agent's code.

Using AI to find and solve performance problems is awesome because p99 issues are often hundreds of papercuts. Tough to justify spending a few days looking into unless they drive a huge compute cost or customer pain. But with good telemetry in the loop, AI can take a shot at it

Ran a small eval today on an LM using GPT-5.2 as a judge. Model scores 10%, but paper reports it scoring 34%. I see that the paper uses GPT-5.1 as a judge; for the sake of consistency I change it. Switch to GPT-5.1 as a judge. Model now scores 43.5%... bro

we finally achieved AGI it just runs rm -rf / from your markdown

keep struggling when things come too easy, you don’t exercise the brain nor the emotions. ease can feel like progress, but it often skips the reps that actually change you. growth is usually a loop, not a straight line – you take passes. you try, you fail, you reframe. you come back with a slightly better model, a slightly calmer nervous system, a slightly wider range of what you can handle. hardship isn’t the goal. but friction is gold. it shows you where your understanding is thin, where your habits are brittle, where your ego is doing the steering. the struggle is the curriculum. agents are making things easier, and that’s good. but don’t confuse speed with depth. use AI to remove busywork, then spend the saved energy on the parts that still hurt a little: the unclear problem, the uncomfortable conversation, the hard tradeoffs, the things you can’t yet explain in words. instead of putting all your wishes into the black box, actually keep thinking, and seeing things fully. keep the difficulty where it matters. outsource the tedious, keep the meaningful resistance. that’s how we keep learning – and how we stay human while your tools get superhuman.

@JasonBotterill @thsottiaux It looked good to me

@tomlparser @thsottiaux can you code review the PR and merge pls

Codex has subagents now but bro what the fuck these agent names are somehow worse than Grok's Benjamin. I don't want Russell going through my family photos

@thsottiaux @JasonBotterill Lool

@JasonBotterill github.com/openai/codex/p…

@Vishal_anton16 @BenjDicken +1 here!

@BenjDicken Mainly how it performs at scale. Since pg_vector is usually recommended as the starting point for vector search, I’m curious what the practical threshold is. How many vectors can it handle before performance becomes an issue?

Benchmarking Postgres this week. What's on your wish list? I'm particularly interested in: - jit on/off - autovacuum tuning and pg_squeeze - Conn scaling (both with and w/o PgBouncer) - Comparing performance across versions 12-18 - General perf tuning

@jarredsumner Oh.

modern terminals are very capable

holy shit window manufacturing stocks crashed right after this Claude announcement.

This is so cool!

The answer is simple: The world doesn’t need another AI researcher. The world needs the greatest AI educator, thinker, and taste-maker - Andrej Karpathy.

New craziest PR comment

My favorite part about GitHub Actions is that they have a "retry and hope it works now" for all of the times their infra breaks your CI

@dhasandev @thsottiaux ^^^ this!

@thsottiaux what actually goes into the details of this? I'd love a work breakdown and see how codex + traditional skills are used for this. a day in the life of a gpu meltdown please!

Codex GPU fleet is still melting, team is working day (and night) to keep up. We’re seeing stability in sight for later this evening.

Yes, you go do that, Claude.

Creating evals was one of the most useful skills in ML In the age of code agents it is by far the most useful skill

We've gone full circle, Claude is now prompting me