AI Coding
AI coding tools and agents, coding-model comparisons and leaderboards, and how AI changes the way developers build software.
The State of AI-Generated Code in 2026: Adoption Soars, Security Lags
AI now generates much of the world's new code and adoption is near-universal, but roughly 44% of AI-generated code carries a security vulnerability, and the EU AI Act's biggest deadline just landed.
Gemini 3.7 Flash Goes Half-Price as Alibaba Open-Sources a 2.4T Model
Gemini 3.7 Flash's half-price launch and Alibaba's open 2.4T Qwen3.8 look like a value-tier win, but the discount expires January 1 and the open model needs 72 GPUs. Read the catches.
An Open-Weight Model Just Cracked the Top 3 of the WebDev Coding Leaderboard
Kimi K3 is the first open-weight model in the top 3 of Arena's WebDev board, as Opus 5 takes #1 and Grok 4.6 undercuts everyone on price.
Everyone Feels Faster With AI. Almost Nobody Can Prove It.
84% of developers feel more productive with AI. Only 20% of their organizations measure whether it's true. That gap between feeling and evidence is becoming a business problem.
GitHub's HydraFusion Beats Claude Opus 5 on One Benchmark of Three
HydraFusion landed in Copilot CLI on September 4 as a research preview. The headline is frontier quality. The benchmark table says the real story is cost.
Claude Code's Weekly Limits Go Up 25% and Down 17% on September 14
Both of Anthropic's numbers are correct. Which one you feel depends on your baseline. Here is the arithmetic, and what to change before the 14th.