A rater that sells nothing it measures
The AI ecosystem has reached its GitHub moment: millions of skills, thousands of MCP servers, endless prompts and agents. Supply is infinite. What's scarce — almost non-existent — is trustworthy intelligence about which of them actually work, for whom, under what conditions, at what cost.
Every marketplace so far optimizes for the wrong thing: SEO, downloads, stars, listings. Anyone can build another directory tomorrow. Discovery has become a commodity. Intelligence has not.
A skill is not software. It's closer to a handbook, an SOP, a playbook — expertise. GitHub evaluates code well. Nobody evaluates expertise. That is the opening.
So tashan isn't another directory that ranks by who shouts loudest. It measures — the rank is did it get adopted, was it kept, was it replaced, did it work, not who paid to be there. When trust is an output of evidence rather than a badge, it can't be gamed and it can't be scraped.
The name. 他山 (tashan) comes from the proverb 他山之石，可以攻玉 — a stone from another mountain can polish your jade. We bring no thumb to the scale — we judge the field with its own public evidence, and let the jade show itself.
What we're building
Start narrow and honest: track the whole MCP field and score it on public signal — adoption, upkeep, freshness, and an LLM read of the actual expertise (deep work vs. thin wrapper vs. slop), live on the Index today. Then layer the things still no one else measures: retention and churn from git history, controlled evals we run ourselves, compatibility across models, and real cost. Capability by capability, the evidence base compounds into something no listing can copy.
The company we want to be
Closer in spirit to the Michelin Guide than to a directory — a verdict people trust because it's earned, not sold. The long arc: the measured layer for the entire AI ecosystem — skills, MCP servers, prompts, agents, models, workflows.
Principles
- Measured, not claimed. If we can't derive it from evidence, we don't publish it.
- Transparent by default. Every signal is defined and reproducible. See the Methodology.
- Ranked on evidence. You can't buy your way up the board — no sponsored slots, no bought scores. The rank is what the evidence says.
- Humble about the unknown. We say plainly what we can't yet measure — and treat that as the roadmap, not a secret.