Updated daily708 verified stories
Research News tracker.
Material AI papers, studies, datasets, evaluations, and scientific advances with direct links to the underlying research.
- 30stories this week
- 708sources cited
- 708verified timeline entries
Verified sourcing
Every story links to the strongest available evidence: original sources when available, otherwise clearly attributed reporting.
Plain English
What happened, why it matters, and what to watch — without the jargon.
No filler
When the signal is thin, we publish nothing rather than padding the feed.
More stories
9 storiesPolicy
Law schools restrict AI use as students enter a changed profession
PYMNTS, citing a Financial Times column, reports that the University of Chicago and UC Berkeley law schools are tightening rules on AI and student devices.pymnts.comInnovation
Martin Cid reports a disputed AGI claim for OpenAI’s Astra
Martin Cid Magazine reports that OpenAI framed Astra as AGI while the benchmark’s reference evaluation scored it 62.7%, far below OpenAI’s reported 99.9% result. The account has not been independently confirmed from the supplied source.martincid.comInnovation
A free chatbot found a six-die solution in about 13 minutes
New Scientist reports that a free chatbot produced a mathematically valid six-die construction that a researcher checked, though the design is impractical and the model’s method remains unknown.newscientist.comSecurity
UK study finds major evidence gaps in open-source AI cybersecurity
A UK-commissioned review of cybersecurity literature on open-source software and open-source AI identifies significant evidence gaps, especially around upstream governance of open-source AI.gov.ukInnovation
Top OpenAI researchers use more than $7,000 a day in AI coding tokens
Business Insider reports that OpenAI’s most intensive researcher users consume more than $7,000 a day in estimated AI-coding token costs, while the company says coding agents are reshaping internal research.businessinsider.comInnovation
Benchmark finds coding agents struggle to build real-world service agents
A new benchmark evaluates whether coding agents can construct complete customer-service agents under realistic business constraints. The paper reports that the strongest tested configuration passed 23.9% of simulations, compared with 82.2% for an expert-authored reference.arxiv.orgInnovation
TipRanks reports cost-focused legal benchmark for Meta’s Muse Spark 1.3
TipRanks reported that Vals AI’s proprietary benchmarks placed Meta’s Muse Spark 1.3 (Max) near Fable 5 and GPT 5.6 Sol while costing 4x–8x less, according to a Vals AI LinkedIn post. The results and pricing claims have not been independently confirmed.tipranks.comInnovation
OpenAI details how coding agents are accelerating internal research
OpenAI says coding agents now generate 3.1 agent-workdays of effort for each human workday in its research organization, while stressing that human researchers still set priorities and judge results.
openai.comInnovation
AI-assisted self-representation wins a Fair Work Commission case
Switzer reports that Macquarie University academic Gregory Baker used multiple paid AI agents while representing himself in a Fair Work Commission dispute. The case shows how AI may lower some legal-access barriers, but it does not establish that AI can replace lawyers.switzer.com.au
One useful briefing each week
Keep up with AI without living in the feed.
Get the week’s verified AI news, original data, useful tools, learning picks, and fresh AI jobs.
Reach people who are learning AI
Hiring an AI professional or launching a useful AI product? Put it in front of people who came here to learn and act.
Post an AI jobSubmit an AI tool