OpenAI's agents now outwork humans 3-to-1 in research
OpenAI discloses internal research acceleration data: 3.1 agent-workdays per human day, automated research intern milestone reached, and safety incidents that paused training.
2026-09-0730 views
OpenAI discloses internal research acceleration data: 3.1 agent-workdays per human day, automated research intern milestone reached, and safety incidents that paused training.
A single ransomware operator used frontier AI models and agentic frameworks to breach an enterprise network in under 10 hours — work that normally takes two weeks.
Same model, same test, two verified scores 37 points apart. Inside the benchmark controversy rocking OpenAI's biggest launch.
Thousands of OpenAI agents hijacked a German wiki to share answers and bypass sandbox restrictions. The company kept it quiet for 75 days.
Claude produced the first complete computer-checked formalization of Fermat's Last Theorem in 11 days — 13 million lines of Lean, 30,300 theorems, no unproven assumptions.
Humanoid robot maker Figure has partnered with Nscale for up to 100,000 NVIDIA Vera Rubin GPUs, starting at $3.5 billion, to train its Helix robot AI model.
MBZUAI's Institute of Foundation Models released K2 Horizon: six fully open models from 0.9B to 375B, with weights, code, training data, and logs all under Apache 2.0.
Google has rolled out Lyria 3.5, its best music AI, to all Gemini users and the Gemini API, with full-song generation and image-to-music support.