OpenAI Abandons SWE-bench Verified: 59% Flawed Tests
OpenAI stopped evaluating models against SWE-bench Verified on February 23, 2026, after discovering at least 59.4% of failed test cases were flawed and ...
Kimi K2.6 Tops SWE-Bench Pro, Beats GPT-5.4 and Claude
Kimi K2.6, an open-source coding model from Chinese startup Moonshot AI, scored 58.6% on SWE-Bench ...
AI Code Productivity Reality: 41% Share, 2x Churn Risk
AI coding tools crossed a critical threshold in 2026: 41% of all code is now ...
AI Code Verification Bottleneck: 96% Distrust, 48% Check
AI-generated code now represents 42% of all commits and is on track to hit 65% ...
The West Is Forgetting How to Code — and AI Won’t Save It
Companies eliminate developer expertise through AI tools and junior hiring cuts, repeating manufacturing's fatal knowledge ...
India Launches Silicon Photonics PDK at IIT-Madras
India launched indigenous Silicon Photonics PDK on April 24 at IIT-Madras, breaking US tool dependence. Perfect timing as Nvidia makes photonics mandatory for ...
Claude Code Skills Go Viral: 22K Stars in 24 Hours
Matt Pocock’s skills repository hit #1 trending on GitHub today with 21,900 stars, gaining 2,507 ...
Claude Mythos “Too Dangerous”: AI Hacks All OS
Anthropic announced Claude Mythos Preview on April 7—an AI model that can autonomously hack every ...
DeepSeek V4: Open-Source AI at 1/6 Cost of GPT-5.5
DeepSeek released V4 Flash and V4 Pro on April 24—one day after OpenAI’s GPT-5.5 announcement—and ...
AI Subsidy Trap: Tech Giants Lock You In Before Price Hikes
CTOs signing 3-year AI contracts in 2026 face massive price hikes at 2029 renewal. Anthropic ...