A crash-handler loop in the Codex/ChatGPT desktop app writes about one dump per second to ~/Library/Application Support/Codex/Crashpad/pending with no pruning, and an Oct 17 report on build 26.1007 measured ~323 GB across ~4 million files. Check that folder: emptying pending/ recovers the space, though it can refill until OpenAI ships a cap.
Cline was sending max tokens instead of max completion tokens , so models like gpt 6 sol and gpt 6.1 sol failed on OpenAI and Azure OpenAI compatible endpoints. The release also makes Claude Sonnet 5.5 the Cline provider's default again when no model is picked…
OpenCode told affected users someone got into a legacy Console system it was decommissioning, and that OpenAI, Anthropic, and other provider keys saved in their workspaces may have been exposed; it tied the Oct 17 breach to a "large scale distillation attack" and says only users with linked external providers who got its email were hit. If you got that email, revoke and reissue those keys in each provider's dashboard and check recent usage, since some users report charges they didn't make…
Point any MCP client at https://ampcode.com/mcp to get two tools: puck , which routes ideas to the right orb and checks on your Amp threads, and orbs , an interactive view of running orbs with portal links…
Each model gets one hour in its lab's first party CLI (Claude Code, Codex, Antigravity CLI, Muse Code), with no internet, to write a C++ Protoss bot. GPT 6 Astra ranks first and GPT 6.1 Sol second at about a tenth of the cost, while Gemini 4 Argon, Claude Opus 5.5 and Claude Fable 5.1 tie at 37 to 39…
Released Oct 16 under the OpenMDW license, Dory 1.0 4B reruns a shared 20 layer block a chosen number of times, so one checkpoint trades speed for accuracy with no model swap; NVIDIA's own tests put it at 50.3 on SWE bench Verified with 1 loop and 61.8 with 4. It has a 256K English context and ships in BF16, MXFP8, NVFP4, and Q5 K M GGUF…
Spotted something we missed? Start a thread.