All news
10603
Nepali Text "Weighed" Three Times More Than English — Tokenizer Fix Couldn't Fully Solve It
Researchers found why models like Llama-3 and Qwen3 spend several times more tokens on Nepali text than on English, fixed the tokenizer's regular expression — but…
SAP researchers teach AI agents to stop flip-flopping between different solutions to the same task
LLM agents with access to tools often behave differently each time they're asked to repeat the same task. A new method called RSD measures how consistent an agent's…
Microsoft Research Unveils Magnetar: a zkVM 10–39x Faster Than Alternatives
Microsoft Research releases Magnetar: proofs of Rust, Go and C program execution, 10–39x faster than rivals, with 12 ms verification and security that rests only on…
Apple Is Bringing AI Chat to Pages, Keynote, and Numbers in macOS 27
Lead Claude Code developers leave Anthropic as the product tops $2.5 billion in revenue
The $448 Billion Paradox: Why 95% of Corporate AI Pilots Deliver Zero P&L Impact
EmbodiedLab: robots tested on experiments, and the best one solves less than half
S3AC brings the critic back to RL for LLMs: gains from 1.5B to 397B parameters
Shanghai AI Laboratory researchers trained a value critic to work with long chains of thought. S3AC beat GRPO-X on models from 1.5B to 397B parameters and held up…
Skill∞: an agent improves its own skills and its own meta-skill, 60.7% on SkillsBench
NUS researchers trained a meta-agent to optimize skills and rewrite its own optimization procedure. Across 83 SkillsBench tasks it scored 60.7%, versus 56.4% for…
China's 27B AREX-2 Outperforms a 1.6 Trillion-Parameter Model
BioStack raises $9.5M for medical AI training data
OpenAI Says Its AI Disproved a 30-Year-Old Theoretical Computer Science Conjecture
XREAL launches 1 Pro+: on-device 2D-to-3D conversion at 60 frames per second
OpenAI Releases GPT-6.1 Terra: Astra-Level Computer Control at 1/10 the Price
OpenAI has introduced GPT-6.1 Terra, a model built for fast computer-use tasks. According to the release, it matches Astra on computer-use performance while costing…
Action Intelligence launches Chorus: robot training data grows 40x
Chinese startup Action Intelligence has unveiled Chorus, an engine that converts video from smart glasses, UMI rigs, and body-worn cameras into training data for…