Share prompts, not code: a reusable prompt suite for evaluating small models (4B-36B)
yaroslavvb · x · 2026-10-27
yaroslavvb argues we should share prompts instead of code — when a better model ships, just rerun the prompts. He published the full prompt suite he used to evaluate small language models in the 4B-36B parameter range (small-models-showdown on spacesheep.dev), reusable for quick head-to-head evals whenever new models drop.
More from coding & agent
- Open-sourced Claude Skill automates pixel VFX layering for game spells — Promptmethus · 2026-10-27
- PSA: Fake GitHub AI tool repos use fake commits and hidden download redirects — Justify_87 · 2026-10-27
- Agent Town: a virtual town where you can drop your AI agents to socialize — johnnypen2445 · 2026-10-27
- FrontMCP adds hot reload for MCP servers: 0.5s restarts, no Claude Code reconnects — DavidAntoon · 2026-10-27
- Forcing personal agents to explain tool calls makes opaque background work auditable — _AustinCalvert_ · 2026-10-27
- A single prompt replaces six-figure competitive analysts with weekly AI monitoring briefs — alex_verem · 2026-10-27