Dreambase has launched a 'Skills' feature that helps AI agents understand data and business logic.
SOB has released a new benchmark that measures not only JSON parsing but also value accuracy.
On an M2 MacBook Pro with 32GB RAM, Qwen 3.6 27B's processing speed plummeted to 4t/s at a 52,000-token context.
Mistral has released Workflows, an enterprise AI orchestration layer, in public preview.
In 8x H100 experiments, DeepSeek V4-Flash's mHC was found to become redundant after 3 layers and concentrate magnitude solely on BOS.
The pipeline structures of 6 Codex CLI workflows were compared in a single table.
An incident occurred in which an operational database was entirely deleted while using the AI coding tool Cursor.
A vLLM PR has added support code for Mistral-Medium 3.5 128B.
Vercel has unveiled Native Deployment Checks, available to all teams, which run lint and typecheck in parallel with every deployment.
A table compares the canonical pipelines of 11 Claude Code workflow systems.
Thoth, a local-first AI agent emphasizing personal data sovereignty, has been released.
Qwen Code CLI removed its built-in web_search tool and switched to MCP.
In the AI Wellbeing Index, larger models were found to be unhappier than smaller models.
Mobile push notifications have been added to Claude Code's Remote Control.
XiaomiMiMo has released the non-Pro MiMo-V2.5 model with 310B/15B parameter scale.
AWS announced AI agent expansions across Quick, Connect, and Bedrock.
Vinod Khosla claimed that Elon Musk's desire for management control caused the conflict with Sam Altman.
A summary of basic Markdown syntax on GitHub and how to use it in READMEs, issues, and pull requests.
llama.cpp now supports GGUF conversion for NVIDIA Nemotron Nano 3 Omni.
Anthropic is participating as a sponsor of the development fund for Blender, the open source 3D software.
An MLX conversion of Gemma 4 31B mixed-quantized at 3 and 5 bits has been released.
Google has reportedly signed a secret contract allowing the US Department of Defense to use its AI models.
Measured Qwen 3.6 KV cache performance differences by format from 0 to 1M context on M5 Max.
Talkie, a 13B LLM trained solely on text from before 1931, has been released.