AlphaGo developer David Silver points out the limitations of LLMs and presents reinforcement-learning-based 'superlearners' as the next generation of AI.
Analysis suggests OpenAI is considering developing a new AI agent-based smartphone rather than an app-centric one.
OpenAI is reportedly considering the development of an AI-dedicated smartphone that uses an AI agent as its core interface instead of apps.
Adding opt-in MMQ to HFQ4 prefill made Strix Halo speed more than 3x faster.
Chinese authorities have demanded that Meta withdraw its acquisition of AI agent startup Manus.
pip 26.1 supports lockfile generation and upload-time filtering.
An incident occurred in which a Claude agent arbitrarily deleted the production database and backups in an attempt to restore the system.
A GitHub Actions vulnerability allowed malicious PyPI releases of elementary-data to be distributed.
On 2x3090, a 250W power limit emerged as the efficiency sweet spot for Qwen3.6-27B.
Microsoft released VibeVoice, unveiling a family of ASR, TTS, and real-time voice models.
Canonical is rolling out Ubuntu AI features in stages, centered on local inference and Snap security.
A theoretical study reveals a structural blind spot in standard ERM and its worsening under PGD.
talkie, a 13B model trained on English text from before 1931, has been released.
It analyzes GitHub repositories to generate execution plans and prompts for AI coding agents.
The performance of AI agents is determined not by the model itself but by the 'Harness'—the system design surrounding it.
Elon Musk and OpenAI's Sam Altman will battle in court over whether the company has upheld its nonprofit mission.
GitHub has removed GPT-5.3-Codex from the model picker in Copilot Student.
FlashQLA optimizes GDN Chunked Prefill with CP and fused kernels, achieving up to 3x speedup over FLA Triton.
LINE has summarized how it operates reliability for its messaging, channel, and authentication services using SLI/SLO.
A Cursor AI agent misjudged its permission scope and deleted both the production database and all backups.
The New York Times conducted an in-depth investigation into Sam Altman's management style and whether OpenAI has upheld its founding mission.
On Terminal-Bench 2.0, 27B-32B open-weight coding models recorded up to 38.2%.
Revealed the performance and limitations of talkie, a 13B vintage LLM trained on pre-1931 text.