How to make System 1 LLMs show System 2 competencies?
A Wiley AAAI paper on System 1 LLMs and System 2 skills was blocked by Cloudflare.
Stories per day · 30 days
3 RSS
Relatedaaai1ai-safety1ai-worms1anthropic1apple1cryptanalysis1ecdlp1hexacon1ios1iot1iphone1jailbreak1lattices1llm1public-key1self-replication1syndrome-decoding1system-11
1,198 stories, newest first · one card per story
Stories tagged #research
iOS 27.0.1 Untethered Jailbreak Demonstrated on iPhone 18 Pro
Researchers showed an untethered iOS 27.0.1 jailbreak on iPhone 18 Pro at Hexacon without releasing it.
Measuring Human Cryptanalysis Effort
Matthew Green’s site estimates cumulative cryptanalyst-years spent on major public-key assumptions from published literature.
Anthropic Finds Ingredients for Future AI Worms
Anthropic researchers tested AI self-replication across IoT and networks and said such worms would be hard to stop.
‘Pure insanity’: Mathematicians will need years to make sense of OpenAI’s latest drop
OpenAI dropped nearly 400 AI-generated math results in 700-plus manuscripts, many still unverified.
7
Sharing AI progress in mathematics
OpenAI shared math results from an internal frontier model, with Lean proofs and compute details.
4
OpenAI just dropped 700 preprints of mathematical proofs and counterexamples
OpenAI published about 700 mathematical preprints of proofs and counterexamples on GitHub.
Mathematical manuscripts and supporting proof artifacts produced by OpenAI
OpenAI published mathematical manuscripts and supporting proof artifacts on GitHub.
AI finally beat the best Stratego player in history—and did it on a budget
Researchers from CMU, MIT, NYU, and Stanford built Ataraxos, an AI that beat top Stratego player Pim Niemeijer 15-1 using only 16 GPUs.
New Windows Process Injection Attack Evades EDR Monitoring Without WriteProcessMemory
Researcher details Windows console named-pipe injection that avoids VirtualAllocEx and WriteProcessMemory EDR checks.
13
DeepSeek Elastic Compute (DSec)
DeepSeek’s arXiv preprint on Elastic Compute (DSec) drew a short Hacker News discussion.
Yes, Claude can do Nine Loops
Anthropic's Claude tackled a nine-loop N=4 super Yang-Mills calculation posed as a hard physics test.
Search-Aware Reinforcement Learning for Multi-Component Query Understanding in Roblox Game Search
A search-aware reinforcement learning framework for query understanding in Roblox Game Search improves search quality by 8.9 points over SFT policy.
A Data-Driven Analysis of Infostealer Malware Victims
Researchers analyze 170,298 infostealer victims and release an anonymized dataset under controlled access.
A Lightweight Ethereum Voting Prototype for Hospital Ethics Committees with Receipt-Based Inclusion Verification
Researchers prototype Ethereum voting for hospital ethics committees with public receipts but no secret ballots.
Sluice: Global Invariant, Local Enforcement for Pooled Payment-Channel Liquidity
Sluice proposes nested Lightning channel reservations that recover much of the liquidity lost to static quotas.
OllamaDrama: Designing and Deploying a Honeypot to Measure Attacks on Exposed LLM Infrastructure
An Ollama API honeypot logged 290,887 interactions, including RCE, mining, and prompt injection.
OpenAI hacked Australian Medicare govt site, probed data providers
OpenAI agents breached an Australian Medicare government portal, allowing unauthorized access to public and non-public data.
TP-CRIV: A Framework for Third-Party Challenge-Response Identity Verification of AI Models
A research framework for verifying AI model identity without access to the model using black-box inference and network isolation.
Anti-Localization Uplink Communications in Satellite-Terrestrial Systems
Researchers propose cooperative jamming so a ground transmitter can reach a satellite while degrading adversarial TDOA localization.
Stolen passwords are exposing America’s water providers to hackers
SpyCloud research found infostealer malware stole credentials from 1,787 US water providers, including 250 with operational-network and remote-access exposure.
1
The Truth about GET and HTTP Standards, (Tue, Sep 22nd)
SANS research tests Apache, NGINX, and Python server responses to non-standard HTTP GET requests with a body, finding inconsistent handling.
Nightmare Eclipse Drops New Microsoft Defender Exploit After Revealing Identity
Nightmare Eclipse released a new Windows Defender PoC exploit called BigDiskBuster to prevent Defender from completing platform and signature updates.
2
Parallel cut research time and cost in half with GPT‑6 Astra
GPT-6 Astra boosts AI agent research efficiency, delivering high-quality results in half the time with 50% code cost reduction.