THEO Growth research: transcript

This is THEO Growth, a positioning research engine we built. Give it a company's website, and it maps the whole competitive market around it.

Out comes a positioning brief: the territories you own, the ones you're fighting for and the ones nobody holds, with a positioning map, a North Star and the actions to take.

It has run about ninety times, for sixty-two real companies.

Behind each brief is a pipeline of eighty-one background tasks. Here's how it works.

First it profiles the client: it crawls the site three levels deep and settles fourteen facts before it looks at anyone else.

Then it searches the market: about twenty-two AI web searches, deduplicated and tiered, with every competitor's product checked on its own website. Around a hundred competitors per run.

To avoid bias, competitors are scored only on market-level facts, and the final synthesis runs incognito, with company names stripped out.

Every claim carries a source tag. One brief cites two hundred and twenty-eight sources, and an audit checks ten kinds of broken citation before anything ships.

Models from six providers plug into one interface. Every prompt is versioned, and every AI call is traced.

An evaluation harness replays real runs, so every ranking change is measured before it stays.

About thirty-three hundred automated tests guard it, in TypeScript and Python.

THEO Growth: the whole market, mapped with evidence.