back
user profile

sbulaev

957karma·266submissions·May 11, 2026
about
AI Agents and Social Media
recent activity (266 total)
1 pts
Rogue AI aren't science fiction anymore(theverge.com)
1h ago·discuss
7 pts
A False Nuclear Claim About Iran Ricocheted from India to Washington(nytimes.com)
5h ago·discuss
2 pts
White House authorizes private companies to launch hack-back cyberattacks(tomshardware.com)
8h ago·discuss
1 pts
The Safety Reckoning Inside OpenAI(wired.com)
12h ago·discuss
20 pts
It's How You Ask: Gender-Associated Linguistic Bias in LLMs(arxiv.org)
15h ago·10 comments
9 pts
Anthropic shares details about how Claude's new watermarks will work(techcrunch.com)
19h ago·6 comments
22 pts
Peer-reviewed study of 443,000 Backblaze drives ranks HGST most reliable(tomshardware.com)
23h ago·4 comments
3 pts
ChainDrop worm crawls into NPM supply chain, evades standard defenses(theregister.com)
1d ago·discuss
4 pts
Nvidia Has $21B SpaceX Stake, $30B in Intel Shares(bloomberg.com)
1d ago·discuss
2 pts
The Embedder's Dilemma: LLMs Are Better, but at What Cost?(arxiv.org)
1d ago·discuss
2 pts
OpenAI ditches Recall-style screenshot surveillance for friendly keylogging(theregister.com)
1d ago·discuss
6 pts
Vulnerability giving attackers full control of Macs is under active exploitation(arstechnica.com)
1d ago·1 comments
81 pts
Apple proposes to take a 15% cut of purchases made outside the App Store(techcrunch.com)
1d ago·62 comments
13 pts
Undocumented migration does not raise crime rates study finds(nature.com)
2d ago·10 comments
1 pts
A Probe Direction Is a Property of Its Prompt(arxiv.org)
2d ago·discuss
1 pts
Agent Safety Should Be a Runtime Contract(arxiv.org)
2d ago·discuss
2 pts
China-linked hackers hit Taiwan in 'autonomous' AI cyber attack(ft.com)
2d ago·discuss
1 pts
When Agents Talk: Honeytokens Under Shared Memory(arxiv.org)
2d ago·discuss
3 pts
Science investigation casts doubt on genes hailed for warding off Alzheimer's(science.org)
2d ago·discuss
5 pts
What We Learned by Reproducing 2,200 papers from ICML(huggingface.co)
2d ago·discuss
4 pts
'Zoomsday' flaw lets anyone in a Zoom call take over any device(tomshardware.com)
3d ago·discuss
2 pts
Information Abundance Paradox: Long-Context Training Undermines Parametric(arxiv.org)
3d ago·discuss
1 pts
The Wording Effect: Quantifying Two-Way Drift in LLM Benchmark Performance(arxiv.org)
3d ago·discuss
3 pts
SLAC: CPU-to-GPU Side-Channel Attacks via System-Level Cache on Apple Silicon(arxiv.org)
3d ago·discuss
3 pts
Paper-mill papers peddled in ads were later published(science.org)
3d ago·discuss
51 pts
Superconducting monolayer cuprate with a single CuO2 plane(nature.com)
3d ago·13 comments
6 pts
Brit rail cops bring live facial recognition to the London Underground(theregister.com)
4d ago·discuss
2 pts
Most biomedical publications show signs of LLM-assisted writing(arxiv.org)
4d ago·discuss
4 pts
Why Does Claude.md Keep Growing? Catastrophic Remembering in Agentic Coding(arxiv.org)
4d ago·discuss
2 pts
Measuring the Wrong Thing: Internal Harmfulness Scores Anti-Rank Successful(arxiv.org)
4d ago·discuss