// Dev
AI doubled merged code. Review is where it piles up, LinearB's data says
Across 2.7 million pull requests, AI-assisted changes were 2.6 times bigger and waited five times longer for review. LinearB now wants to feed that context to agents.
// DEV
Dev
Tools, agents and languages: what changes for people who write code.
// Dev
Claude's API can now hold the conversation for five minutes, cutting long-context latency
Message threads, in public beta, let agents send only the new turn instead of the whole history. The token bill doesn't change; time to first token does.
// Dev
Your AI shopping agent can't get past a CAPTCHA, and some stores like it that way
Amazon blocks rival agents on purpose, Walmart says its checks catch Meta's Muse by accident, and banks warn about giving agents your password.
// Dev
ChatGPT folds Work into the main chat in December, and Codex starts sharing your allowance
OpenAI is merging the Chat and Work tabs and moving paid plans to one token-based allowance across ChatGPT and Codex. Instant chat stays off the meter.
// Dev
Claude will no longer rebuild your favorite game, even if you rename it
Anthropic added instructions steering Claude away from recreating existing games amid a wave of AI-assisted fan ports of Halo, Call of Duty and GTA V.
// Dev
Your agent lost the response. Do not let it retry the refund.
A benchmark found agents claimed success in 90% of cases where they had actually done something twice. The fix is boring distributed-systems hygiene.
// Dev
A reporter with zero coding experience vibe-coded his own Muse Charm for $50
The Verge's Sean Hollister bought a Meta-approved ESP32 devkit and let Claude Code turn it into a talking, Tamagotchi-style AI pendant. He has no idea if the code is any good.
// Dev
One shared ChatGPT link could take over your account, thanks to a math keyboard
Researchers chained a postMessage bug, an innerHTML sink and a YouTube endpoint to escape ChatGPT's HTML preview sandbox. OpenAI fixed it and paid $5,000.
// Dev
Someone put a NES CPU inside a 4B-parameter model. Mario runs. Slowly.
Drew Breunig built a working 6502 emulator into the weights of Qwen3.5-4B, with Claude Opus 5.5 doing much of the legwork.
// Dev
Factory's Droid 1.0 makes scripted tool calls the default for coding agents
The model-agnostic harness posts 65.2% on Factory's own Terminal-Bench 4 runs with Claude Opus 5.5. Details are still thin.
// Dev
Anthropic's Claude extension on Cursor shipped an SSH backdoor for nearly 12 hours
A malicious update pushed through Anthropic's official Cursor marketplace account reached up to 5,000 developers. How the account was taken over is still unexplained.
// Dev
Your AI agents need ID badges and audit logs. Most frameworks don't ship them.
Partnership on AI researchers studied the four most popular agent frameworks and found they can't reliably say which agent did what, or who approved it.