Same model, 13.3% to 38.3%
Nothing about the model changed. The score almost tripled. Same Model, 13.3% to 38.3% OpenAI changed two API settings on ARC-AGI-3. Same model, same weights,...
The approval prompt is not the sandbox
Two harnesses, two very different answers to the same question: what is actually stopping the agent? The Approval Prompt Is Not the Sandbox Codex ships...
Your model file is untrusted input
A 65-byte file can crash llama.cpp. Your Model File Is Untrusted Input Sixty-five bytes is not a model. It is a valid GGUF header with a tensor whose second...
Merged is not shipped
The fix was merged a year ago. Nobody got it. Merged Is Not Shipped: Why a Year-Old Fix Never Reached You A bug got fixed, correctly, a year ago. Then it sat...
Your VPN is fine. iOS swapped the address.
Hey, Here is a failure that survives every reasonable thing you would check. One app on a phone, on cellular, talking to a machine on my own VPN. It pairs....