to the guy guessing API keys on my AI gateway:
78,769 attempts in 82 minutes. peak: 2,030 requests per minute.
every single one got a 401. zero successful authentications, zero tokens, $0 in model spend.
you could've read the first one. rip.
gpt-6.2-astra:
"The main chat mistook the repository’s existing CI setup for a requirement of the MVP.
Three GitHub Actions workflows already had runs-on: @namespacelabs. That explained where those tests were configured to run. It didn’t establish that the Bot Factory needed Namespace to work.
The mistake became explicit when the assistant wrote “Keep Namespace” into its proposed release plan. When you said “Implement the plan,” it carried that assumption forward. Later handoffs even described xxx as the owner’s choice. The visible history shows the assistant putting Namespace into the plan; it doesn’t show a separate instruction from you making it an MVP requirement."
@thsottiaux I was wondering why it started to set up sandboxes on Namespace as the target architecture... 💀
pure token burning here
asked gpt-6.2-astra with reasoning on max to change a default in my nixos config repo. it edits ~/.codex/config.toml and tells me it’s done. git diff is empty.
I have to explain that changing the config in the repo is the whole point of the task.
things like this happen so often that coding becomes grueling 💀 @thsottiaux
running Codex in WSL2 over SSH from the Windows ChatGPT desktop app can get messy fast. one thing that helped: set my custom gateway as the default in WSL2. now plain codex uses it. adding -p loads a profile and skips the shared background process.
Having said that, I’m ready and fully charged to test the hell out of GPT‑6.2 Astra and Sol 🥵
Just look at those benchmarks. Let’s see what our agent setups make of them.
https://t.co/b0VBOqsJeN
We run agents 24/7 at https://t.co/1etEkS5AG0 so we test a lot of models.
Lately, we’ve tested glyph-cluster extensively via @CheaperInfer. Really positive results 👀
It’s #8 in this graphic. A reminder to test on your own workloads, not just follow rankings.
Some of you have been asking very politely. Some of you have just been asking.
GPT-6.2 Astra and Sol are now available.
Super efficient and built to perform where it counts—the real world.
https://t.co/qr2iP0GpfX
We run agents 24/7 at https://t.co/1etEkS5AG0 so we test a lot of models.
Lately, we’ve tested glyph-cluster extensively via @CheaperInfer. Really positive results 👀
It’s #8 in this graphic. A reminder to test on your own workloads, not just follow rankings.
Come build your next app or agent with us in the API.
Eligible ChatGPT Pro and Business subscribers can claim up to $500 per month in API credits.
Get started: https://t.co/IUETYqLqov