{"name":"inference-optimisation-skill","repo":"sybil-solutions/inference-optimisation-skill","url":"https://midorreal.com/project/sybil-solutions-inference-optimisation-skill","repository":"https://github.com/sybil-solutions/inference-optimisation-skill","description":"Agent skill for hill-climbing local LLM serving speed (decode, speculative decoding, prefill, KV, startup, TTFT) on any hardware and quantization","facts":{"license":{"value":"MIT","source":"https://github.com/sybil-solutions/inference-optimisation-skill","checked":"22 Nov 2026"},"language":{"value":"Python","source":"https://github.com/sybil-solutions/inference-optimisation-skill","checked":"22 Nov 2026"},"releases":{"value":"0","source":"https://github.com/sybil-solutions/inference-optimisation-skill/releases","checked":"20 Nov 2026"},"contributors":{"value":"1","source":"https://github.com/sybil-solutions/inference-optimisation-skill/graphs/contributors","checked":"20 Nov 2026"},"archived":{"value":"no","source":"https://github.com/sybil-solutions/inference-optimisation-skill","checked":"22 Nov 2026"}},"stars":1776.0,"writeup":{"what_it_is":"An agent skill for hill-climbing the speed of a local LLM server, covering decode, speculative decoding, prefill, KV cache size, startup time and time to first token, on any hardware and quantization. It describes a workflow of computing theoretical ceilings, changing one thing at a time, measuring against controls, logging results and gating on long-context retrieval.","audience":"Engineers tuning local LLM inference servers on their own hardware","claims":[{"kind":"specific","claim":"Measured decode speedup on Intel Arc B390 for one user","excerpt":"Decode, 1 user: 43.6 → **53 tok/s** (median)","status":"not_checked"},{"kind":"specific","claim":"Measured prefill speedup at 256K context on Intel Arc B390","excerpt":"Prefill 256K: 297 → **628 tok/s** (2.1×)","status":"not_checked"},{"kind":"specific","claim":"Includes a benchmark harness that works with any OpenAI-compatible server","excerpt":"`bench.py` benchmarks any OpenAI-compatible server.","status":"not_checked"},{"kind":"specific","claim":"Works with Claude Code and any agent that loads SKILL.md-style skills","excerpt":"It works with Claude Code and any agent that loads `SKILL.md`-style skills.","status":"not_checked"}],"alternatives":[],"written_by":"AI, from the project's README","date":"2026-11-18"},"why_now":null,"owner_supplied":null,"score":{"verdict":"mid","hype":42,"reality":17,"gap":26,"momentum":null,"confidence":60,"version":"score-2.0","rule":"verdict-2.0","calculated_at":"2026-11-22T07:00:00+00:00","verdict_source":"rule"},"early_read":null,"changes":[{"date":"2026-11-16T18:40:04+00:00","text":"Repository created","source":"https://github.com/sybil-solutions/inference-optimisation-skill"}],"cite":"inference-optimisation-skill is a GitHub project at https://github.com/sybil-solutions/inference-optimisation-skill written mainly in Python, described there as \"Agent skill for hill-climbing local LLM serving speed (decode, speculative decoding, prefill, KV, startup, TTFT) on any hardware and quantization\".","attribution":"Data from Mid or Real, https://midorreal.com/project/sybil-solutions-inference-optimisation-skill, checked 22 November 2026. Attribution with a link is required.","still_real":null}