LLMs Reward Expertise, Not Beginners: What the Data Shows
HN front page today: LLMs reward domain expertise over raw effort. Senior engineers capture 5x more AI productivity gains. Your skills still matter ...
OpenAI Astra Solves 10 Decade-Old Math Problems for $2,000
OpenAI revealed Astra by solving 10 decade-old math problems for $2,000 with Lean 4 proofs. ...
WebMCP in Chrome 149: Make Your Website Agent-Ready Now
Chrome 149’s WebMCP origin trial is live, but zero sites have shipped it. Here’s how ...
Microsoft Project Perception: Agentic Security Is Live
Microsoft Project Perception entered public preview today. Three AI agents find, triage, and patch vulnerabilities ...
GPT-5.6 Luna Drops 80%: Rethink Your Model Routing Now
OpenAI cut GPT-5.6 Luna to $0.20/M input tokens on July 30. Here is what changed, ...
Retyping LLM Code Beats Cognitive Debt — Here Is Why
A 2026 study found unrestricted AI coding groups had a 77% maintenance failure rate. The case for manually retyping LLM code is not ...
Agentic Fatigue: How AI Tools Are Burning Out Devs
New research shows AI coding tools create cognitive overload, not relief. Here is what the ...
DeepSeek V4 Flash 0731: Smaller Model, Better Agent Scores
DeepSeek shipped V4-Flash-0731 on July 31 with vendor benchmarks showing a 13B-active-parameter model outscoring its ...
Open-Weight AI 2026: When Self-Hosting Beats the API
Open-weight AI finally matches closed models. But the self-hosting economics only pencil out above 150M ...
Kimi K3 Open Weights: What Developers Need to Know
Moonshot AI's Kimi K3 — 2.8 trillion parameters and open weights — is live on ...