Research
Anthropic study finds robots can perform three-quarters of US physical tasks, but are cost-competitive for 0.3 percent
A new robot exposure index finds that about 80 percent of work time is exposed to either robots or large language models.
Research
Papers, benchmarks, and findings from labs and universities.
- Research
Anthropic says Zhipu's GLM-5.3 can build cyber exploits on its own and that its safeguards were bypassed in 64 to 100 percent of tests
The Frontier Red Team says exploit-building capabilities like Claude Mythos Preview's are now available in a model released without meaningful misuse protections.
- Research
Hinton, Bengio, Pachocki and Clark co-author a paper warning that automated AI research could trigger an 'intelligence explosion'
More than 20 researchers say AI now writes most of the code at frontier labs, and they urge policymakers to gain visibility into R&D automation.
- Research
Fudan-linked study of 700 task logs finds AI agents did more of the work of building a model, but humans still made the decisions
The team examined how its own researchers and their agents built Atria Dawn Preview, a 744-billion-parameter mixture-of-experts model.
- Research
Stanford and Caltech's HomeBody connects GPT-6 Astra straight to a humanoid robot's skills to tidy an unfamiliar kitchen
The system skips the usual trained control layer. It pairs a swappable vision-language model with spatial memory and a simulated digital twin of the room.
- Research
Cryptanalysts say OpenAI and Anthropic models cracked two long-unsolved Enigma messages
One developer had GPT-6 Astra find, research and decode a message that had defeated researchers since 2005.
- Research
Forecasting study finds leading AI experts badly underestimated how fast the field would move
An interim report from the Forecasting Research Institute says AI reached gold-medal level at the math olympiad five years before the median expert forecast.
- Research
Anthropic says Claude found a CRISPR-like enzyme system in phage DNA whose function is still unknown
The first result from Anthropic's new biology lab is a preprint that has not been peer reviewed. It describes 'array-associated reverse transcriptases' that sit next to runs of repeating DNA.
- Research
OpenAI releases MentalHealthBench, an open set of 1,215 mental health conversations graded against clinician-written rubrics
More than 80 psychologists and psychiatrists from 22 countries wrote 5,262 criteria covering everyday well-being as well as emergencies.