Problem discovery · inside AJO

Discovery.

Discovery refuses to assume the current priority is the right one. Its whole job is to keep reformulating what the fleet should be solving in the first place — the front of the loop, never a step tacked on before the real work starts.

Deep exploration — decompose, scout, converge

exploring

A domain is broken into fully distinct angles, each one scouted in parallel by its own deep web-search loop, then converged into a single verified, cited dossier.

One domain, split into genuinely distinct angles, each scouted in parallel until everything converges into a single dossier — nothing ships without independent citation across multiple sources and a full QA pass.

Discovery exists because building the wrong thing well is worse than building nothing at all. It’s the whole front of the loop, made real as a system rather than a step someone remembers to do — the part that refuses, completely, to assume the current priority is actually the right one. It reformulates how the fleet searches, runs deep exploration across every angle worth checking, and stands up whatever focused engine the search actually needs. The capability itself is the identity here; any single assignment is just today’s task.

01How it works

Two ways to find a problem.

The never-settled search — the lead never assumes the current priority is the goal, full stop; its entire job is to keep reformulating how the fleet looks for what genuinely matters.
Frame a problem — the lead decomposes it fully, a researcher surveys adjacent work, a coder runs real exploration probes, QA validates every piece — and what comes out the other end is a sharp, fully grounded problem statement.
Explore a domain deep — a synthesizer breaks a domain into genuinely distinct angles and dispatches parallel scouts, each one running an iterative web search-and-fetch loop against exactly one angle.
Converge and verify — every scout’s dossier converges into one, and a winning dossier only ships once it carries independent citation across multiple sources and has cleared a full QA pass — never before.
The auditor — it stress-tests the reasoning for rigor and relevance, flagging the moment a convenient priority is being mistaken for a genuinely real fit.

02The engines it runs

Purpose-built, and replaceable.

Discovery was never built to marry one method. It stands up a genuinely focused engine for whatever needs finding, keeps the ones that earn their place on merit alone, and retires the ones that don’t — completely, without sentiment.

Pain Point Pipelinelive

The current engine — one query in, one cited pursue, hold, or pass out.

Sextantretired

An earlier cited-answer search engine, built in full and then set aside once it was clear it wasn’t earning its keep. What worked in it carried forward whole into the Pain Point Pipeline.

More, as neededplanned

Discovery is built to keep standing up laser-focused engines the moment a new kind of search is genuinely needed — each one judged and kept or retired purely on its own merits.

It also leans fully on the Research project’s own engines — the Researcher decides what’s genuinely worth asking, and more laser-focused pipelines get built the moment a new kind of search is needed.

03Why it leads

Aim before build.

Discovery sits at the true front of the whole loop, as a non-negotiable stage rather than an afterthought. What it frames, the Researcher interrogates fully, the Pain Point Pipeline grounds in real evidence, and Tool Engineering builds. Get the aim wrong here and every stage after it is wasted completely — which is exactly why an auditor stress-tests the reasoning, flagging the instant a convenient priority is being mistaken for a genuinely real fit.

Runs inside AJO. Code is private; this page is the record.

← All work