AI
Artificial intelligence: LLMs, agents, diffusion, vision, NLP and the latest from top labs.
593 published articles
Medical AI accountability
OpenAI's health AI claims to outthink doctors. Then its own lead walked it back.
OpenAI's ChatGPT Health launch pits a VP's bold claim against her own team's caution, and a lawsuit alleging dangerous advice.
2026-07-29
Artificial Intelligence
The hardest lesson for AI reasoning engines: when to shut up
MIT researchers propose OS-Pruner, a plug-in that dynamically stops chain-of-thought reasoning when further computation isn't worth the token cost. Tests show 20-60% length reduction with minimal accuracy sacrifice.
2026-07-29
Artificial Intelligence
The fraud detector that got better by ignoring what it was told
PREF-Gate separates fraud detection into a label-free context expert and a label-derived evidence expert, then lets a validation gate decide which to deploy. On Amazon and YelpChi, it drops the evidence entirely; only on TFinance does it accept a small mixture. The paper forces the field to confront whether label leakage, not model architecture, drives many published GNN gains.
2026-07-29
Artificial Intelligence
Clean your AI's input, not its reasoning. The speed gains are absurd.
Researchers at CRIL demonstrate that preprocessing techniques preserving model counts dramatically speed up model sampling, enumeration, and direct access queries when formulas are compiled into d-DNNF circuits. The study tests 1,425 benchmarks and shows that removing defined variables with compatible orderings solves up to 47 more instances than no preprocessing.
2026-07-29
Mechanistic Interpretability
Gemma 4 knows physics well enough to steer it, and that changes everything
New research on gemma-4-E4B-it reveals that the model's internal representations of materials science mechanisms are causally linked to its answers. The study combines multiple probing and intervention techniques to show that physics knowledge is encoded in state transformations, not just static patterns.
2026-07-29
AI Engineering
83 Claude Code tips just became a community wiki, and it reveals the new skill that matters
A community-maintained GitHub repo collects 83 tips for Claude Code, documenting the shift from vibe coding to agentic engineering. The resource reveals context budgeting as the new core skill.
2026-07-29
AI Research
Distributional RL heads make risk claims that are mostly false. A new audit proves it.
A large-scale audit shows that distributional RL agents systematically fabricate risk trade-offs at the very states where practitioners would most trust them. Across QR-DQN, C51, and IQN on MinAtar, zero of the strongest claims were confirmable, and acting on the agents' CVaR advice sometimes performed significantly worse than chance.
2026-07-29
Microsoft AI
Microsoft matches GPT-5.6 in Excel with a cheaper, older-gpu model
Microsoft's MAI model in Excel matches GPT-5.6 for common tasks at lower cost, uses fewer tokens than rivals in Copilot, and runs on H100 and A100 GPUs, lowering the bar for widespread deployment.
2026-07-29
AI Architecture
The Model Race Has a Second Track Now
Anthropic capping Claude Opus 5's cyber skills, Sakana's Fugu orchestrating model swarms, and Claude Code's new multi-model routing all point to the same shift: the next competitive edge in AI systems is orchestration, not just raw model size.
2026-07-28
AI Models
OpenAI shipped a long context fix. It just won't tell you how much better it got.
OpenAI shipped a long context fix without sharing any benchmarks. That silence might tell developers more than the update itself.
2026-07-28
Artificial Intelligence
France's €360M bet on AI: nine clusters to train the next generation
France allocates €360 million to nine AI clusters under France 2030, aiming to boost training and research. Two new training projects receive €11 million to upskill workers in AI-related fields.
2026-07-28
Partnership
ElevenLabs just got the UK government to put a testbed for voice AI in public services
ElevenLabs has signed an MOU with the UK government's DSIT to test voice AI in public services, focusing on accessibility for visually impaired, elderly, and linguistically diverse communities. The company is expanding its London HQ and doubling UK staff to 200.
2026-07-28