Product docs

Enough docs to trust the thing.

Not a novel. Not empty marketing fluff either. This page pulls the sharpest parts of the public GitHub docs onto the site so people can understand what Synthetic Users is, how it works, what exists now, and what the intended runtime looks like.

Synced from GitHub README one-pager Architecture + diagram

Start here Open architecture diagram

Overview

What Synthetic Users is

Synthetic Users is a pre-launch pressure test for product surfaces that usually break only after traffic hits: landing pages, launch copy, onboarding, pricing pages, and rough product specs.

Instead of asking one polite model for feedback, it runs multiple sharp synthetic personas and turns the mess into a short decision memo: what landed, what got missed, what killed trust, what to rewrite next.

How it works

Short loop, high signal

Input

Submit a URL or paste raw launch text, onboarding copy, pricing copy, or a PRD excerpt.

Normalize

The runtime fetches or cleans the input, determines the attack surface, and structures the claims.

Simulate

Four synthetic personas react with different patience, trust thresholds, technical depth, and bullshit tolerance.

Summarize

The output clusters repeated objections into a compact memo with score, verdict, top leak, and rewrite-next guidance.

API / Inputs / Outputs

What the beta runtime accepts right now

Accepted inputs

url

rawText

attackSurface

notes

Returned output

score

verdict

topLeak

memo bullets

persona reactions

Current vs target

What exists now versus where the runtime is headed

Current public artifact

Working beta site with live analysis form, docs page, static assets, and API routes exposed through the Cloudflare worker runtime.

Target runtime

MiroShark-style simulation layered with Aeon-style orchestration so launches can be tested, rewritten, rerun, and compared in a repeatable loop.

Personas + output

The output should be blunt and useful

Synthetic set current beta

skeptical engineer

distracted founder

non-technical buyer

cynical operator

Decision memo format from architecture doc

what landed

what got missed

what killed trust

rewrite next

test after that

System modules

The product architecture without drowning people in boxes

Intake service

Accepts URLs, pasted text, and metadata. Fetches and cleans content.

Surface parser

Classifies landing page, launch copy, onboarding, pricing, or spec and extracts claims and proof blocks.

Persona engine

Stores persona templates and instantiates the run-specific set.

Simulation orchestrator

Runs persona passes, handles batching, retries, and repeatable loops.

Aggregator / scorer

Clusters objections and summarizes clarity, trust, hook, and drop-off risk.

Memo generator

Outputs the short decision artifact and keeps reruns comparable.

Assets + sources

Go deeper only if you want to

Source