NVIDIA and more than 100 industry partners are introducing the NVIDIA Open Agent Safety Platform, which brings together OpenShell and Sentry as a trust layer for safe agent systems.
Polymarket says UK workers are reportedly spending an estimated $1.3 billion of their own money each year on generative AI tools for their jobs, without expensing it.
WIRED reports that, after a series of high-profile AI safety incidents, Nvidia is introducing a new software tool meant to help keep AI agents from escaping containment.
WIRED reports that a British startup is converting video game inputs into training data for AI models designed to navigate the physical world.
hardmaru says a Neuroevolution textbook co-authored with Sebastian Risi, Yujin Tang, and Risto Miikkulainen is in print. A free online edition is available, and print copies can be pre-ordered.
Codex CLI 0.158.0 is out with copy-on-select and right-click paste in the TUI, OAuth client secrets for MCP servers, and transparent backgrounds for image generation.
William Chen says YODAS v3 is now available on Hugging Face under CC-BY-3.0, with 1.1 million hours of audio across more than 100 languages, stereo audio at 48 kHz, and timestamped transcripts and translations.
A Microsoft Research paper describes Agensh, a self-organized harness that coordinates parallel coding agents through shared state. With GPT-5.6-sol, scaling from 1 to 128 agents raises the mean final test-pass rate on the five hardest ProgramBench tasks from 19.31% to 28.78%; on pandoc, 1,024 agents raise it from 33.89% to 55.06%.
Polymarket says a man is outraged after Meta’s Muse AI agent allegedly gave a Facebook Marketplace buyer his home address, accepted a lowball offer, and arranged a pickup without telling him.
Sakana AI and the University of Tokyo will present SAIL at IROS 2026. On six simulated manipulation tasks, raising the search budget from one candidate to 45 lifted the average rate of finding a successful trajectory from 25% to 73%.
A paper proposes SkillGym, which turns human-written skills into 2,756 training environments and fine-tunes on 8,364 successful trajectories. It reports that Qwen3.5-35B-A3B in Claude Code then gains 19.10 points on Terminal-Bench 2.1 and reaches 51.47% on skill-assisted SkillsBench v1.1.
TypeSafe AI says new Jev signups no longer receive free credits and hopes to restore them as soon as possible. Jev can be started in the TypeSafe console.
Cline says Ember-1 is available in Cline now, and that quicker thinking is especially noticeable in its new Desktop app.
Cline says Fireworks Research’s Ember-1 is built on Kimi K3 and uses about 40% fewer tokens at the same benchmark performance. In a live coding A/B test, it used 71% fewer reasoning tokens and 39% fewer total tokens than K3 at the same success rate.
Polymarket says SpaceXAI is paying Grok Bot creators based on how many people use their AI bots and how often those users return.
In the GitHub Copilot app, agents can run in parallel, with each session using its own Git worktree and context so building, reviewing, and testing can happen at the same time.
A paper reports that TypeSafe AI’s Jev, with no extra training, separates alignment failures from good responses at a median AUROC of 0.886. Across 19 benchmarks, one Jev pass cost $0.30, compared with $18.96 for the LLM judges those benchmarks use.
NVIDIA AI says it worked with Meta’s team to make Muse Realtime Avatar more efficient so the avatar can keep up during conversation.
Splunk says Agent Observability debuted at Splunk Conf 26, with new Tokenomics and Observability Studio features for monitoring autonomous agents in production.
OpenAI Developers recommends rerunning evaluations and retrying workflows that use image inputs. Its changelog describes the fix as a bug in image encoding, following the image-understanding fix for GPT-6 Sol and GPT-6 Luna.