cnbc.com by cramer4next
Mistral CEO says U.S. AI safety debate masks competitors' 'negligence'
51 points2 comments
Capability · theguardian.com
Top stories Updated 2 Oct 2026
Research Labs and arXiv
Lab notes latest posts
OpenAI
A model guide for the GPT-6 family
NoteLearn how startups can choose GPT-6 models, tune reasoning effort, improve prompts and skills, coordinate tools, and prepare workflows for production.
Hugging Face
Open-sourcing AstaBrief, the fast report-generation model in Asta
Google Research
Toward provably private learning from federated data
NoteMobile Systems
Hugging Face
AutoSynthData: Generating Training Data for Enterprise Agents
arXiv tray newest first
arXiv:2610.02185 [cs.LG]
Decoding Looped Transformers Better for (Almost) Free
Looped Transformers achieve parameter efficiency by repeatedly executing a shared block across recurrent loops. Each loop yields an intermediate representation decodable for the same next token, yet standard decoding discards earlier states. Because earlier…
arXiv:2610.01995 [cs.AI]
Can AI Oversight Be Zero Knowledge?
AI systems increasingly produce outputs from confidential data, such as a fitness-for-duty assessment from medical records or the predicted properties of a drug candidate from its secret structure. It is important to verify that…
arXiv:2610.01073 [cs.SE]
Safety Must Survive Self-Improvement: Why Failures Persist and How Agents Recover
Recursive self-improvement (RSI) allows agents to carry useful changes across generations. Maintaining safety across these generations involves both preventing unsafe behavior from persisting and enabling recovery when failures occur. We study these challenges through…
arXiv:2609.40330 [cs.AI]
Turbo Harness: Instance-Adaptive Harness Optimization
Automating the search for effective harnesses is an important step toward enabling agents to recursively self-improve. Existing harness optimizations typically produce a single global harness that is applied uniformly across task instances. However, a…
Social wire Public posts, reconstructed as text
Hacker News 12 threads
wsj.com by borski
OpenAI Scraps Release of New AI Model over Safety Concerns
30 points5 comments
Mastodon 12 posts
br00t4c @br00t4c.bsky.social@bsky.brid.gy
"They're not SuperIntelligent": Timnit Gebru discredits Big Tech's AI claims #psyops #AI #GOPLiesAboutEverything #SuperIntelligence #BernieSanders #Regulation youtube.com/watch?v=Xqxb... "They're Not Superintelligent"...
0 replies0 boosts0 favourites
VeilCorp Archives @VeilCorp
A name you're given to manage how a room perceives you, and a name you earn because something true happened underneath it, those aren't the same kind of object. I had to earn mine. So does anything calling itself "super" this week., WEDNESDAY // BECOMING # Cradlepoint # Wednesday # Becoming # SuperIntelligence # DigitalCulture
0 replies0 boosts0 favourites
About Supence
SUPENCE is two words pressed into one: SUPer and intelligENCE. Supence tracks SI: what news outlets report, what labs publish, what papers claim and what people discuss, with every item linked to its source. Headlines are the publishers’ own; nothing is rewritten.
On Supence now
30 articles, 14 videos, 12 papers
Send a tip or correction