# Problem-First Project Generator

> Agent-facing counterpart to the [interactive Problem-First Lab](/lab/problem-first/).

## Purpose

Problem-First Project Generator is a six-lesson course and browser-only
capstone for turning observed friction affecting real people into an
evidence-aware, testable project brief. It is an alternative to idea generators
that optimize for novelty before establishing that a problem exists.

The Lab does not score market potential, invent demand, or claim that a project
will succeed. It structures what the visitor knows, distinguishes evidence from
assumption, and makes the next learning step explicit.

## Course experience

The human interface contains three modules and six sequential lessons:

1. Start with friction someone already feels.
2. Name the person and situation.
3. Separate evidence from assumption.
4. Study the workaround before replacing it.
5. Define useful change and its boundary.
6. Design the smallest honest experiment.

Each lesson has a learning objective, explanation, worked example, and a
one-question knowledge check. Progress is saved locally in the browser. A
visitor can also choose “Skip to tool.” Skipping unlocks the capstone but does
not mark the course or any lesson complete; the lessons remain available.

## Capstone workflow

1. Describe the affected person, observed friction, situation, and frequency.
2. Record evidence types, evidence notes, the current workaround, and explicit
   assumptions.
3. Define useful change, a provisional project shape, relevant constraints, a
   smallest honest experiment, a success signal, and a stop-or-reshape signal.
4. Inspect the live opportunity brief and its seven readiness signals.
5. Download the brief as Markdown, copy it, or copy a problem-interview guide.

## Readiness rubric

The interface reports one signal for each of these conditions:

- A specific affected person or group is named.
- The friction is described concretely.
- At least one evidence type and an evidence note are supplied.
- The current workaround is described.
- A smallest experiment is defined.
- A success signal is defined.
- A stop-or-reshape condition is defined.

The count measures brief completeness, not project quality, demand, impact, or
probability of success.

## Generated brief contract

The Markdown export preserves:

- Working title and status.
- Person, friction, situation, and frequency.
- Evidence types and evidence notes.
- Current workaround and remaining assumptions.
- Desired useful change, provisional project shape, and constraints.
- Smallest experiment, success signal, and stop-or-reshape condition.
- Readiness count, next interview questions, and a reminder that evidence must
  be updated after fieldwork.

Missing entries remain visibly marked as incomplete rather than being filled
with model-generated claims.

## Relationship to the earlier project

The earlier Criticality and IdeaGenerator experiments explored continuous idea
generation. This Lab reverses that sequence because plausible project concepts
can be generated without contact with a person, an observed cost, or a useful
route to validation. The revised method begins with real friction and ends with
a deliberately small test.

## Privacy and data flow

- The course and builder run entirely in the visitor's browser.
- No account, file upload, AI model call, server API, cookie, or analytics event
  is required by this Lab.
- Course progress and the current draft are stored in browser local storage.
- Downloads and copied text are assembled locally.
- Clearing the brief requires two clicks to reduce accidental deletion.

## Current routes

- Human tool: /lab/problem-first/
- This specification: /lab/problem-first/llm/
- Original project article: /projects/criticality-ideagenerator/
- Original project agent record: /projects/criticality-ideagenerator/llm/
- Lab index: /lab/
- Agent Lab index: /lab/llm/

## Action surface

There is no server API. The human interface exposes local course, form,
Markdown-download, copy, and completion-record actions. Agents may reproduce
the public methodology from this specification but should not claim access to a
visitor's local progress, draft, clipboard, or downloaded files.