GPT-5.5 Codex Keeps Cutting Its Own Reasoning Off at Exactly 516 Tokens
A GitHub analysis of 390,195 coding-session responses found GPT-5.5 disproportionately cuts off its own reasoning at exactly 516 tokens, a pattern likely caused by a batching bug rather than an intentional change.
A Flask Creator Says Anthropic's Newest Models Got Worse at Using Tools
Flask creator Armin Ronacher found that Anthropic's newest models, Opus 4.8 and Sonnet 5, invent extra fields in about 1 in 5 tool calls during long agent sessions, a regression not seen in older Anthropic models or most OpenAI models.
Claude Code Users Report Other People's Data Showing Up in Their Sessions
Two new GitHub issues describe unexpected data appearing in Claude Code sessions, with one confirmed case of another user's live server credentials leaking in and being used without authorization.
An AI Agent Screened 2.4 Million Crystals and Found Four New Superconductors
An AI system called ElementsClaw screened 2.4 million candidate crystal structures and flagged four new materials that a lab has since synthesized and confirmed are genuine superconductors.
Three Popular Ways to Train Reasoning AIs Turn Out to Be One Formula
A new proof shows that three widely used reinforcement-learning recipes for training reasoning models - GRPO, Dr. GRPO, and DAPO - are all just different operations on a single number, the spread of rewards within a group of sampled answers.
Why Asking an AI the Same Question 10,000 Times Barely Helps
A new analysis shows that sampling many answers from an AI and picking the most common one hits a hard ceiling because the samples are correlated, not independent, so thousands of extra tries can be worth only a couple of genuinely new ones.
Samsung's Trick Makes a Tiny 4B Agent Nearly Match a Model 18 Times Bigger
Samsung R&D UK and Queen Mary University of London published DuoMem, a distillation method that took a 4-billion-parameter agent from a 4.3 percent task-success rate to 77.9 percent, nearly matching a 72-billion-parameter teacher model's 87.1 percent.
Microsoft Starts a 2.5 Billion Dollar Company Just to Get Businesses Using AI
Microsoft launched a new business called Frontier Company, backed by a $2.5 billion investment and staffed by roughly 6,000 people, whose sole job is deploying AI inside client organizations rather than just selling software or consulting on it.
OpenAI Is Reportedly in Early Talks to Give the US Government a 5% Stake
OpenAI is reportedly in early discussions about the US government taking roughly a 5% stake in the company, worth about $42.6 billion at its current $852 billion valuation, ahead of its planned September 2026 IPO.
Cloudflare Now Lets Sites Block AI Training Bots While Keeping Search
Cloudflare has split its AI bot controls into three separately toggleable categories - search, live agent, and training crawlers - free for all sites, with training and agent bots blocked by default on ad-supported pages starting September 15, 2026.
NVIDIA Starts Taking a Share of Its Cloud Partners' Revenue, Not Just Selling Them Chips
NVIDIA announced a new arrangement where it earns a share of the cloud revenue its partners generate from NVIDIA-supported data center capacity, on top of its usual chip sales, with first partners Sharon AI and Firmus building campuses totaling hundreds of thousands of GPUs.