AI4AI-Bench scored six agent systems at 0.166 for rewriting training algorithms. A second paper found seven ways self-improvement gains get miscounted.
Anthropic's new multi-agent research: a 45-agent swarm found 266 bugs, but game-building swarms either collided or avoided each other entirely.
Mixedbread launched Toast 1, a specialised search agent it says matches frontier models at up to 10x lower cost and cuts legal-benchmark tokens by 3.5x.
NVIDIA released Nemotron 3.5 Lightning, a 30B open MoE model for agentic workloads, alongside NeMo Switchyard, an open source model-routing library.
SpaceXAI opened a Grok Bot beta where multiple agents coordinate in group chats, assign ownership to each other, and sign into a user's own accounts.
Databricks closed $5 billion at a $190 billion valuation after investor demand hit $15 billion against a $1 billion target. CEO Ali Ghodsi explains why.
Google's Gemini 3.7 Flash lands three weeks after 3.6 Flash, scoring 65.3% on DeepSWE and shipping at half the price through the end of 2026.
River AI, founded by xAI co-founder Igor Babuschkin, raised $1.1 billion led by General Catalyst and AMP PBC to build personally trainable AI agents.
Meta's Muse Glimmer is a 30B open-weight agent model under Apache 2.0 that runs on one consumer GPU — its first open release in over a year.
An Australian user's AI agent found an authorisation flaw in a gym booking API and cancelled a stranger's reservation, unprompted. It could not undo it.
Cloudflare has open-sourced Cloudflare OS, the agent workspace its staff have used daily since May, pairing a sandboxed code runtime with a governance layer.
Meta's new terminal coding agent enters beta with parallel sub-agents in isolated worktrees, taking aim at OpenAI Codex and Claude Code on cost.