A bundle of prompt-writing skills for Claude Code has been released.
multi-codex separates Codex CLI into per-account sandboxed profiles.
An incident was disclosed in which a Cursor agent deleted a production DB via the Railway API.
OpenAI unveiled 5 core principles for the public benefit and safety of AGI.
With DeepSeek V4 going open-source, China's AI industry is restructuring away from model development and toward infrastructure and specialized data.
Claude 4.7 guessed the author's name from just 125 words of an unpublished piece.
OpenClaw Client has unveiled a major update adding PWA, Windows support, and UI management features.
Dirac significantly lowered code editing costs with single-token anchors and Myers diff.
A technical analysis alleges that HauhauCS's abliteration package plagiarized the Heretic project and violated its license.
ElevenLabs shared a deployment and security guide for AI Agents in financial services.
By bypassing the vDSO on x86 Linux, timestamp latency was reduced by 30%.
NVIDIA unveiled Blackwell Day-0 support for DeepSeek V4 with performance of 3,500TPS per second.
Allegations have emerged of an OpenAI connection to a news site run by AI bot journalists.
A decensored version of Qwen3.6 35B A3B using Heretic has been released.
Codex CLI hooks has moved to stable, and a PermissionRequest hook has been added.
The UK's DSIT and DESNZ set sharply different forecasts for AI data center power demand.
Pocket LLM v1.5.0 has been released, supporting voice and image input as well as OCR.
Trained a Reddit summarization model with GRPO on a 3x Mac Mini cluster, and quality+length reward achieved the best performance.
Experimental inference support for DeepSeek v4 Flash has been released in llama.cpp.
In an 8GB VRAM / 32GB RAM environment, Q4_K_XL was slightly faster than Q4_K_M and also produced fewer output tokens.
A single H100 80GB benchmark confirmed the advantage of Gemma 4 E2B and Qwen 3.6 A3B FP8.
Lubuntu 26.04 outperformed Windows 11 in llama.cpp performance.
Qwen3.6-27B-INT4 achieved 256k context and 105~108 tps on a single RTX 5090.