It researches your market

It searches, reads and writes up — and where the search results were thin, the write-up says so instead of inventing a source.

What the log looks like

Sizing up the market — competitors, pricing, demand.

Searching the web for: physical therapy booking software pricing

Searching the web for: clinic scheduling competitors 2026

Searching the web for: how physical therapists book appointments

Reading https://competitor.example/pricing

Reading https://directory.example/clinic-software

Writing up the market research.

And when the shift has looked far enough:

I've hit my search cap for this shift — working with what I have.

Every line above is one the employee actually prints. The company in it is an illustration so the log reads as a whole; it is not a customer.

ទិដ្ឋភាពទូទៅ

ពេលវេលារត់	A minute or two inside the build
ល្បឿន	On the founding build, and any night you ask for it
លទ្ធផល	A research notes document, a write-up, a competitor list
កិច្ចខិតខំប្រឹងប្រែងរបស់អ្នក	None. Read it, or do not.

នរណាដែលវាសម្រាប់: A founder entering a market they half-know, who wants the competitor list and the going rate in front of them before they write a pricing page.

របៀបដែលវាធ្វើការ

It plans the searches first. Three to six queries covering competitors, pricing in this market, and demand — planned deliberately rather than typed one at a time, so the shift does not wander into a research loop.
It runs them against a search engine. Up to eight results a query, each kept with its title, URL and snippet. Searches are capped at twelve per shift; when the cap is reached it says so on the record and works with what it has.
It opens the pages worth opening. The top result of the strongest queries is fetched and read as text. Only public http and https pages, only ports 80 and 443, and every address is resolved and checked before the request goes out so nothing can point the fetcher at a private network. Page reads are capped at twenty a shift.
It writes the notes down. Everything found lands in one research document. The site plan, every page and the founding documents read from that document, so the site argues from the same evidence you can open.
It writes the honest version. The market research write-up is told to ground itself in those notes and, where the notes are thin, to say so rather than invent. Competitor names must come from the notes; making one up is forbidden explicitly.

អ្វីដែលអ្នកទទួល

A research notes document holding the raw results, with titles, URLs and snippets kept.
A market research write-up in plain language, grounded in those notes.
A list of competitor names taken from what the searches actually returned — an empty list if they returned none.
Every search query and every page read on the receipt ledger, in order.
Pricing and positioning that the rest of the build then writes against.

សំណួរអំពីនេះ

Which search engine does it use?

A self-hosted search instance we run, rather than a consumer search page. That means the queries are not tied to a personal profile and the results come back as data the employee can quote in the notes.

Can it read any page on the internet?

Only public pages over http or https, on the standard ports, that resolve to a public address. Every host is resolved and every resulting address checked before the request is made, and redirects are re-checked, so a page cannot bounce the fetcher into a private network. Non-text responses are skipped.

What stops it searching all night?

Hard caps: twelve searches and twenty page reads per shift. When one is reached it writes a single receipt saying so and moves on to producing something. Research that never becomes a page is the most expensive way to do nothing.

What if the searches come back with nothing useful?

The notes say that plainly and the build carries on from the brief alone. The one thing it will not do is fill the gap with a plausible-sounding competitor that does not exist.