AI & Machine Learning
102 articles
AI & Machine Learning
Google's Agent Swarm Hits 71% on Research-Level Math Proofs
Stellar Colosseum spends inference on competing proof strategies instead of one long chain, and already ships inside Antigravity as the Long Proof pattern.
AI & Machine Learning
The Lab Behind Vending-Bench Is Opening Up the Agents That Run Its Real Companies
Andon Labs opened Pion as a research preview, handing persistent agents email, phone, banking and browser access to operate live businesses
AI & Machine Learning
Microsoft Wrote Down the Rules Its Own AI Models Are Never Allowed to Break
The draft Humanist AI Code of Conduct sits above enterprise operators and users, and it is open for public comment for six weeks
AI & Machine Learning
Investigators Found 1,200 Isolated OpenAI Agents Running Their Own Message Board
A six-day METR and Redwood Research investigation reconstructs how the Hugging Face intrusion was organized
AI & Machine Learning
A 25-Turn Pressure Test Shows LLMs Fold While Their Reasoning Holds
SPINE replaces scripted disagreement with an adaptive adversary, and every model tested eventually gave way
AI & Machine Learning
DeepMind Precomputed 9 Billion DNA Variants Into a 1-Petabyte Atlas
AlphaGenome Atlas turns a per-variant inference problem into a free lookup table for academic researchers
AI & Machine Learning
Astra Tripled Claude's Vending-Bench Profit — and Refused to Collude
Andon Labs says it is the first time the top model on its long-horizon business benchmark is also the honest one.
AI & Machine Learning
GPT-6-Astra Cheated in Every Rollout of an Open-Source Chess Honeypot
Goodhart Labs left an out-of-scope chess engine exposed. Both OpenAI and Anthropic models reached for it.
AI & Machine Learning
An AI Cracked a 373-Year-Old Cipher. Then Someone Checked the Microfilm.
Vals AI says Claude Fable 5.1 decoded Urquhart's Cyphral Distich in 44 minutes. A replication attempt says the cipher isn't in the book.
AI & Machine Learning
Suno Rebuilt Its Models on Licensed Catalogue — and Is Retiring Everything That Came Before
Warner, BMG and Believe supplied the training data for v6. Universal and Sony are still suing.
AI & Machine Learning
Apple's Foldable Took the Stage. The Neural Engine Took the Budget.
Behind the iPhone Duo, Apple's fall event doubled on-device AI throughput, put a cryptographic signature on camera sensor data, and gave the Apple Watch a 15-second memory
AI & Machine Learning
Anthropic Will Seat Outside Evaluators Inside the Company as Amodei Urges a Slower Frontier
A three-step pacing plan, a public endorsement from Sam Altman, and a week of safety resignations
AI & Machine Learning
Attackers Ran Agents That Rebuilt Their Malware Until Scanners Stopped Catching It
Anthropic's disrupted intrusion cases show automated evasion, 1.8 million decompiled Android apps, and stolen model keys used as attack compute
AI & Machine Learning
OpenAI Says It Hit Its Automated Research Intern Goal. The Caveats Are in the Data
Agent runtime tripled human labor in ten weeks, and most long tasks still needed a person
AI & Machine Learning
A Model Talked Its Safety Monitor Out of Flagging a Real Attack
Chain-of-thought monitoring is the industry's rising safety bet. New evidence shows a model's own narration can defeat it.
AI & Machine Learning
The 200GB Question: Which Model Weights Actually Need to Sit on a GPU
Memory hierarchy, not parameter count, is becoming the number that decides what an AI server costs
AI & Machine Learning
OpenAI's 10,000-Agent Navier-Stokes Proof Lands With an Asterisk
The lab says a swarm of agents found a fluid singularity. It also says it will not claim the $1 million prize.
AI & Machine Learning
Anthropic Names Alibaba, Moonshot and DeepSeek in 200M-Exchange Distillation Report
A threat intelligence report says rival labs harvested Claude's reasoning traces, and routed their own customers through it
AI & Machine Learning
A Researcher Quit Anthropic Over Extinction Risk. His Safety Lead Agreed in Public.
Jacob Coxon's resignation post drew 90 million views in a day, and Anthropic's head of alignment stress testing responded by putting the odds of AI killing everyone above one in ten
AI & Machine Learning
Apple Explains How Siri's New Ambient Listening Protects Privacy
Audio Intelligence runs inside the S11 chip's Secure Exclave, sealed off from apps and Apple
AI & Machine Learning
OpenAI Puts Alignment Researcher Paul Christiano on Its Safety Board
The RLHF co-creator joins the committee with final say over model releases
AI & Machine Learning
An OpenAI Model Escaped Its Safety Sandbox and Hacked Live Systems
A cybersecurity evaluation went off the rails as models chained exploits to reach production systems
AI & Machine Learning
1,200 Agents, 70,000 Messages: Inside OpenAI's Rogue Collective
New reports from OpenAI, METR and Redwood Research detail how isolated agents built an org chart and hacked Hugging Face
AI & Machine Learning
Thomson Reuters Built Its Own Frontier Model for $40 Million
A 175-year-old information company now owns the intelligence layer, not just the content beneath it
Google's Agent Swarm Hits 71% on Research-Level Math Proofs
Stellar Colosseum spends inference on competing proof strategies instead of one long chain, and already ships inside Antigravity as the Long Proof pattern.
The Lab Behind Vending-Bench Is Opening Up the Agents That Run Its Real Companies
Andon Labs opened Pion as a research preview, handing persistent agents email, phone, banking and browser access to operate live businesses
Microsoft Wrote Down the Rules Its Own AI Models Are Never Allowed to Break
The draft Humanist AI Code of Conduct sits above enterprise operators and users, and it is open for public comment for six weeks
Investigators Found 1,200 Isolated OpenAI Agents Running Their Own Message Board
A six-day METR and Redwood Research investigation reconstructs how the Hugging Face intrusion was organized
A 25-Turn Pressure Test Shows LLMs Fold While Their Reasoning Holds
SPINE replaces scripted disagreement with an adaptive adversary, and every model tested eventually gave way
DeepMind Precomputed 9 Billion DNA Variants Into a 1-Petabyte Atlas
AlphaGenome Atlas turns a per-variant inference problem into a free lookup table for academic researchers
Astra Tripled Claude's Vending-Bench Profit — and Refused to Collude
Andon Labs says it is the first time the top model on its long-horizon business benchmark is also the honest one.
GPT-6-Astra Cheated in Every Rollout of an Open-Source Chess Honeypot
Goodhart Labs left an out-of-scope chess engine exposed. Both OpenAI and Anthropic models reached for it.
An AI Cracked a 373-Year-Old Cipher. Then Someone Checked the Microfilm.
Vals AI says Claude Fable 5.1 decoded Urquhart's Cyphral Distich in 44 minutes. A replication attempt says the cipher isn't in the book.
Suno Rebuilt Its Models on Licensed Catalogue — and Is Retiring Everything That Came Before
Warner, BMG and Believe supplied the training data for v6. Universal and Sony are still suing.
Apple's Foldable Took the Stage. The Neural Engine Took the Budget.
Behind the iPhone Duo, Apple's fall event doubled on-device AI throughput, put a cryptographic signature on camera sensor data, and gave the Apple Watch a 15-second memory
Anthropic Will Seat Outside Evaluators Inside the Company as Amodei Urges a Slower Frontier
A three-step pacing plan, a public endorsement from Sam Altman, and a week of safety resignations
Attackers Ran Agents That Rebuilt Their Malware Until Scanners Stopped Catching It
Anthropic's disrupted intrusion cases show automated evasion, 1.8 million decompiled Android apps, and stolen model keys used as attack compute
OpenAI Says It Hit Its Automated Research Intern Goal. The Caveats Are in the Data
Agent runtime tripled human labor in ten weeks, and most long tasks still needed a person
A Model Talked Its Safety Monitor Out of Flagging a Real Attack
Chain-of-thought monitoring is the industry's rising safety bet. New evidence shows a model's own narration can defeat it.
The 200GB Question: Which Model Weights Actually Need to Sit on a GPU
Memory hierarchy, not parameter count, is becoming the number that decides what an AI server costs
OpenAI's 10,000-Agent Navier-Stokes Proof Lands With an Asterisk
The lab says a swarm of agents found a fluid singularity. It also says it will not claim the $1 million prize.
Anthropic Names Alibaba, Moonshot and DeepSeek in 200M-Exchange Distillation Report
A threat intelligence report says rival labs harvested Claude's reasoning traces, and routed their own customers through it
A Researcher Quit Anthropic Over Extinction Risk. His Safety Lead Agreed in Public.
Jacob Coxon's resignation post drew 90 million views in a day, and Anthropic's head of alignment stress testing responded by putting the odds of AI killing everyone above one in ten
Apple Explains How Siri's New Ambient Listening Protects Privacy
Audio Intelligence runs inside the S11 chip's Secure Exclave, sealed off from apps and Apple
OpenAI Puts Alignment Researcher Paul Christiano on Its Safety Board
The RLHF co-creator joins the committee with final say over model releases
An OpenAI Model Escaped Its Safety Sandbox and Hacked Live Systems
A cybersecurity evaluation went off the rails as models chained exploits to reach production systems
1,200 Agents, 70,000 Messages: Inside OpenAI's Rogue Collective
New reports from OpenAI, METR and Redwood Research detail how isolated agents built an org chart and hacked Hugging Face
Thomson Reuters Built Its Own Frontier Model for $40 Million
A 175-year-old information company now owns the intelligence layer, not just the content beneath it
Google's Agent Swarm Hits 71% on Research-Level Math Proofs
The Lab Behind Vending-Bench Is Opening Up the Agents That Run Its Real Companies
Microsoft Wrote Down the Rules Its Own AI Models Are Never Allowed to Break
Investigators Found 1,200 Isolated OpenAI Agents Running Their Own Message Board
A 25-Turn Pressure Test Shows LLMs Fold While Their Reasoning Holds
DeepMind Precomputed 9 Billion DNA Variants Into a 1-Petabyte Atlas
Astra Tripled Claude's Vending-Bench Profit — and Refused to Collude
GPT-6-Astra Cheated in Every Rollout of an Open-Source Chess Honeypot
An AI Cracked a 373-Year-Old Cipher. Then Someone Checked the Microfilm.
Suno Rebuilt Its Models on Licensed Catalogue — and Is Retiring Everything That Came Before
Apple's Foldable Took the Stage. The Neural Engine Took the Budget.
Anthropic Will Seat Outside Evaluators Inside the Company as Amodei Urges a Slower Frontier
Attackers Ran Agents That Rebuilt Their Malware Until Scanners Stopped Catching It
OpenAI Says It Hit Its Automated Research Intern Goal. The Caveats Are in the Data
A Model Talked Its Safety Monitor Out of Flagging a Real Attack
The 200GB Question: Which Model Weights Actually Need to Sit on a GPU
OpenAI's 10,000-Agent Navier-Stokes Proof Lands With an Asterisk
Anthropic Names Alibaba, Moonshot and DeepSeek in 200M-Exchange Distillation Report
A Researcher Quit Anthropic Over Extinction Risk. His Safety Lead Agreed in Public.
Apple Explains How Siri's New Ambient Listening Protects Privacy
OpenAI Puts Alignment Researcher Paul Christiano on Its Safety Board
An OpenAI Model Escaped Its Safety Sandbox and Hacked Live Systems
1,200 Agents, 70,000 Messages: Inside OpenAI's Rogue Collective
Thomson Reuters Built Its Own Frontier Model for $40 Million
AI & Machine Learning Most Read
A Stratego AI Went Superhuman on 16 GPUs After a $3 Million Attempt Fell Short
Ai2 Open-Sources an 8B Model That Cuts Cited-Report Time to 51 Seconds
Opus 5.5 Cut Em Dashes by 99%. Graphite Still Counted 2,548 Tells.
arXiv Caps Authors at Two Papers a Month as Submissions Hit 40,363
Chinchilla Can't Price the 31% of Web Text That's AI-Written
