Programming

The Axiom project verified CFSG in Lean with AI agents assisting a 24-day sprint— a watershed for proof assistants and formal mathematics.

Parallel ETH refutation claims from OpenAI and Anthropic kick off an organized review—what developers should know before changing assumptions.

Moonshot's K3.1 Pro targets coding, 3D, and finance with a high-speed variant; builders are watching the November 4 weight drop and OpenRouter pricing.

Compare ClosedXML, EPPlus, and ExcelDataReader with examples, pitfalls, and security notes.

Anthropic and Theorem showed LLMs can produce assembly-level proofs and uncover long-hidden bugs in production software.

Critical PHP RCE flaws triggered global emergency patching; here's how engineering teams should upgrade runtimes, redeploy, and verify estates after AI-assisted disclosure.

Microsoft Research's MAI-Base-2 mixture-of-experts release challenges closed labs with a 2.8-trillion-parameter open-weight checkpoint and tooling aimed at enterprise fine-tuners.

Genkit Dart 1.0 brings Firebase-aligned agent flows to Flutter. Stackademic on APIs, on-device models, and building agentic mobile apps.

Researchers at Harbin Institute of Technology report SparseEngine delivers up to 2.5× the serving throughput of vLLM on agentic inference patterns—if sparse attention tricks translate to production.

Anthropic and GitHub bring Haiku 5.5 to Copilot in October 2026 with lower cache-read pricing and subagent workflows for multi-step coding.