CTO @ https://xitter.cf/t.co/Lh8ZUVvHpR. Created https://xitter.cf/t.co/LSsWcUU86S Host of @readkubelist
Austin, TX
Joined July 2009
- Tweets4.2K
- Following842
- Followers1.1K
- Likes4.2K
i'm official at the point where ai code review is quicker than ci... and ci is pretty short on this project
time to focus on ci now?
waiting for ci.
waiting for tests locally
waiting for code review
doesn’t matter what it is. waiting sucks
oh no. i’ve been using jev and matched our output to this disclosure. ugh
despite the unfortunate timing, we take disclosures seriously at @typesafeai
at approximately 6:49PM this Friday evening, we became aware of a massive unauthorized leak of ALL customer noul outputs
in effort to maximize transparency, here is the link: gist.github.com/diogo149/810…
sometimes i think 6.1 sol is a troll. it just told me:
"Marc, I'm not hand-writing this code, I'm using a model to generate it."
the only thing worse then not having websockets continuously updating a page is having them update SOME parts of the page (e.g. the GitHub PR page)
i have to keep state: this part refreshes and that part doesn't
what is wrong with apple developer connect. this would be not possible without ai agents. one of the worst flows possible
maybe we don't need to System 1 model everything? but we still will
DeepSeek V4.1 Flash 1017 runs at 1,521+ tok/s on Fireworks Ultrafast.
Fast enough to ship Tetris in 11 seconds, 7x faster than Claude Opus 5.5.
Join the Ultrafast waitlist: bit.ly/3UA0wKq
in 10 hours today, i managed to use an entire week of the $500 codex plan. on sol 6.1 medium (ultra fast)
but it was worth it. building something pretty cool
i'm at 81% weekly codex quota left, but only a few hours into the week. I have a usage reset expiring, tried to use it and got this. @thsottiaux why can't we use our resets whenever we want?
seriously. just get daybreak access from @OpenAI and scan your old code that’s still running
it won’t catch everything but “future you” will come back and say thanks.
🚨 CRITICAL: GitLab flaw (CVSS 10.0) affects 12 years of releases.
CVE-2026-110264 could allow unauthenticated remote code execution on self-managed servers.
Read - thehackernews.com/2026/10/cr…
i spend a lot of time talking to an LLM.
i realize now that many times i am chatting with a human i’m actually chatting with their LLM.
inefficient proxy
i'm enjoying the new System One models. typesafe/jev & clef and now @CloudflareDev has Clairic.
Been using a lot of jev, and now I'm able to replace a slow LLM call that takes a collection of linear stories, and have Clairic just extract the relevant data for me. Quickly.
