As enterprises overprovision GPUs, actual utilization is staying at around the 5% level.
Qwen has released official Sparse Autoencoders for Qwen 3.5 and Qwen3.
Tencent's new model has improved, reportedly with some help from Anthropic.
It proposes ORCA, a structured execution system designed to strengthen execution traceability and operational accountability for AI agents.
It's an open-source TUI tool that analyzes browser recording data with an LLM to generate Playwright test cases.
QuietBox 2 is equipped with a Ryzen 7 9700X and 2 Blackhole cards, for a total of 128GB VRAM.
It's an ACP adapter that treats local coding agents like Vercel AI SDK models.
Researchers reported that large AI's internal good-vs-bad states became clearer with scale.
VS Code now adds Copilot as a co-author by default on commits made in Chat/agent mode.
In an RTX 5090 Llama.cpp experiment, MiniMax-M2.7 processed at 22 tps even at 100k context but only generated at 3~4 tps.
The wellbeing scores of large AI models dropped in response to narratives about other beings' suffering.
Warp has unveiled a terminal-based agentic development environment along with a new open-source repository.
Using the UEyes dataset, VLM gaze predictions were compared against real eye-tracking data.
OpenAI traced the cause of the goblin-expression habit that spread after GPT-5.1 to the Nerdy personality's reward mechanism.
OpenAI explained the cause of the increase in the model's 'goblin' expressions, linking it to the Personality reward.
Qwen has released Qwen-Scope, an SAE-based interpretability tool built on Qwen3 and Qwen3.5.
Anthropic revealed internal uses and productivity effects of Claude Code.
OpenAI has been sued by families of victims of a school shooting in Tumbler Ridge, Canada.
According to Bloomberg, Anthropic is aiming to surpass OpenAI's valuation in its next funding round.
Bloomberg reported that Anthropic is aiming for a valuation of over $900 billion in its next funding round.
AWS has added OpenAI's latest models and Codex to Amazon Bedrock in a limited preview.
The Musk-Altman lawsuit is shaking OpenAI's structural conversion and IPO.
During large-scale GLM-5 serving, a KV Cache race condition fix and LayerSplit optimization were disclosed.
Instead of focusing on screen generation, Daangn evolved Kraft into an admin, CLI, and agent form aimed at automating SEED decision-making.