so uh... just dumped from my codex cli sessions. logged back in. immediately dumped again. I have advanced account security enabled, and the yubikeys work fine. it just kicks me out again.
@thsottiaux@OpenAIDevs@OpenAI
i loved borland.
I started here when I was 11 so I'm not sure if started actually counts.
I also did a little bit on apple iie and c64, but don't recall what those were.
this has been looping in my head all week. asked opus 5.5 for it's own take on a video. @eigenrobot
"the song is the prisoner. every frame traces its actual waveform as a 3D knot on a green vector scope, caged in rotating 4D polytopes (tesseract, 24-cell, 600-cell). each drop shatters the cage and unlocks a color, and the service puts it back in a bigger one. twice"
it occurs to me that I could have just built the case to put the gpus sideways like a blade. it would need additional supports but then exhaust from one never goes anywhere near another one.
v2 i guess
another day, another thing that needs fixing.
before and after is kind of insane though. upsiren phase change on the die, and minus pad pro on the memory took this card from worst to best out of the batch.
i kept getting gpu temp flags, but the core was reading 37c. looked around for memory temp reading tools, found none.
astra: eh, i'll just build one. literally 90 seconds later.
so uh... did daybreak blue access requirements for orgs change? @thsottiaux when did all of these boxes become mandatory? that's gonna make it inaccessible to most orgs, just for blue.
We will give one banked reset for every day you don't have access to Astra on your paid ChatGPT plan, starting today. Team is moving mountains to give access as fast as we can.
First one will land in ~ 3 hours. There is still time to create your account if you don't have one.
Releasing Pokee-Isaac 28B — the world’s first real 10M-token context frontier-class agentic model, deployable on a single GPU (starting from RTX 4090 or equivalent).
New proprietary non-decoder-only architecture:
• 93.3% RULER at 10M tokens
• Up to 137K tokens/s prefill on one B200 with 10M-token context
• Leads BFCL v4 and τ³-bench in our evaluation
• Lowest combined attack success rate among evaluated models on DTAP security red-teaming benchmark
Pricing and deployment:
💰 $0.15/M input · $1/M output
🔒 Deploy in your VPC, on-premises, or on-device, with Day-0 support for @vllm_project and @sgl_project
Technical blog: console.pokee.ai/model
Technical report: console.pokee.ai/pokee-isaac…
API: console.pokee.ai
hey Sol. did you over engineer this a bit? it feels over engineered. look around and clean up the parts that are excessive.
"yeah, I was a little over enthusiastic. cleaning up now."
I've attached a screenshots and a short recording. sometimes I get five or six words in a row. sometimes it stops every two words. in comparison, I just dictated this entire tweet in one go on the same device.
Anthropic had the mandate of heaven and manage to screw it up. Again.
they had openclaw. they had Fable. they had wrestled the default attention of the normies.
no I don't even think about them anymore.