About Assay

What Assay is

Assay (v.) — in metallurgy, the test that determines the purity and value of an ore. You cannot tell gold from pyrite by looking at it. You take a sample, put it through fire, and weigh what survives.

That is the entire business.

The market is filling up with AI agent libraries — collections of prompts and personas, tagged and listed, almost none of them tested. Collecting is cheap. Assaying is the work. Raw material → assay → asset. That's the only sequence we sell.

Who runs Assay

Assay is run by one accountable human and a team of AI agents. The agents do the research, the drafting and the analysis. A human — Anderson — reviews everything and is the one who ships it, signs it, and answers for it. Nothing goes out without that review.

We say this on the front page rather than in a footnote, because it's the thing we're actually selling. An AI team that nobody checks is worth nothing. An AI team with an examiner, a written constitution, and a kept record of its own failures is a different object.

Assay is a brand, not a registered corporation. There is no office, no staff roster, no funding round. There is a method, a record, and a person who signs.

How we work

Every methodology we publish goes through the same three gates:

1. A written constitution. Agents work under rules they can't amend: external actions require human approval, no commitments made on the brand's behalf, every claim must answer "where did this come from," and any normative claim without a source is labelled unverified rather than dressed up.
2. An independent examination. The agent that writes something is never the agent that grades it. The examiner grades against what a senior human professional would produce, not against "seems reasonable," and it is allowed to fail us.
3. A kept record. We keep the failures. The methodology behind our first product failed its first examination at 72/100, was rewritten, and passed the re-examination at 98/100. We keep both reports, and we'll show you both.

To be precise about what that is and isn't: our examiner is our own. This is an internal quality system, honestly described — not third-party accreditation, and we will never present it as one.

What you can expect from us

• We tell you when not to buy. Our first product's page says, in the description: if you have fewer than 5 engineers or no repo history, don't buy it — the sample size makes the diagnosis unreliable. We'd rather say that before checkout than after.
• We label our uncertainty inside the product. Any threshold we haven't field-calibrated is marked as uncalibrated where you'll actually see it.
• No invented evidence. No fabricated benchmarks, no invented case studies, no borrowed testimonials. Where we don't have data, the product says so. We have removed real numbers from published drafts because we couldn't trace them to a first-hand source, and we'd do it again.

What we sell

Self-assessment kits and diagnostic methodology for engineering organisations trying to answer a question adjectives can't: what actually changed when we adopted AI?

Start with the AI Software Factory Maturity Self-Assessment Kit — 4 metrics, a 0–4 level, a 90-day plan, 60 minutes with your own data. Nothing connects to your systems.

Questions, disagreements, or a copy of an examination report: hello@useassay.co