Skip to content

Repository files navigation

The Anchorflight Check for Synthetic Users

The Anchorflight Check for Synthetic Users

Know when synthetic users help, and when they'll fool you.

AI-driven product loops move too fast for traditional validation, creating a dangerous illusion of user certainty. Anchorflight is an interactive triage tool that helps you weigh a product decision. In about two minutes it tells you exactly where synthetic cohorts are highly reliable, and exactly when you must halt the loop and bring in real humans.

Built for product teams (PMs, designers, and researchers) making decisions about pricing, onboarding, and feature gates.

© 2026 Terry M. Patterson · Licensed under Apache-2.0

License Powered by OpenRouter

📖 User Guide with screenshots and a walkthrough of every feature.


About the name

Aviation has always built its vocabulary on maritime tradition: we fly aircraft using knots and nautical miles. Anchorflight intentionally bridges that gap. The flight is your development velocity, the pace at which AI lets you build. The anchor is the empirical human data that keeps every synthetic pass grounded. The framework exists to hold both at once: move fast, but stay tethered to real people where it counts.


Two ways to run a decision through Anchorflight

Both paths share the same goal: knowing when to trust synthetic users. Pick the one that matches how your team builds.

Building continuously with AI → the continuous-AI loop (primary)

If you go from intent to a working screen in hours, with no natural pause to validate, start with the Active Checkpoint. Run one decision through four gates and get an honest verdict:

  1. Triage runs first and sets how hard the exit is. Low-stakes, reversible work gets a loose exit; high-stakes, irreversible work gets a hard one.
  2. Frame runs a cheap synthetic gut-check that surfaces objections and edge cases. This is idea expansion, directional, not validation.
  3. Inspect gives a directional read on what you've built so far, then you decide: refine, reframe, or proceed.
  4. Ship gate hands you a ranked list of exactly what to validate with real humans before you ship, sized to a scarce budget.

Most runs end "go." The few that say "stop" are the ones that matter.

Building in defined stages → the 8-gate framework (alternative)

If you work in phases (discovery, design, build, QA) with handoffs and reviews, use The 8 Gate Anchorflight Framework and the Cohort Due Diligence wizard. Walk a decision through eight sequential gates, from decision clarity and risk to grounding, calibration, and compliance, and get a clear proceed / adjust / stop verdict with a downloadable report.


What It Does

Synthetic AI users can accelerate research, but misused they create feedback loops that distort product metrics, inflate conversion assumptions, and mislead entire roadmap cycles. Anchorflight exists to prevent that by telling you, per decision, where synthetic users are reliable enough to trust and where only real humans will do.

Modules

Module What it does
Active Checkpoint The continuous-AI loop triage. Run one decision through the Triage → Frame → Inspect → Ship gates for an honest go/stop verdict. Start here.
The 8 Gate Anchorflight Framework Visual 8-step pipeline with a gate inspector; click any gate to see its operational guide and lifecycle rationale
Cohort Due Diligence Step-by-step wizard that walks a decision through the eight gates and produces a proceed / adjust / stop verdict with a downloadable report
Synthetic Persona Prompt Architect Generates a grounded, bias-resistant synthetic persona system prompt from real usage data, then lets you interview that persona live and audit for drift
Usability Tester Directional, AI-assisted interface evaluation from a synthetic persona's perspective, with screenshot/image analysis
Lifecycle Rationale Strategic matrix showing where synthetic users are and aren't appropriate across standard and agentic product lifecycles

Take the built-in 90-second interactive tour (first-visit chip, or the Menu) to see the whole loop in motion.


Architecture

Anchorflight is a static front end (a single, self-contained index.html plus a few supporting pages) with no database, no accounts, and no cloud storage. The only server-side component is a lightweight, stateless serverless function (api/ai.js) that proxies AI requests to OpenRouter so the API key stays off the client; it stores no prompt content. All other computation (triage scoring, gate logic, persona and prompt assembly, report generation) runs in your browser.


Usage

No build step. No npm install.

Open locally

git clone https://github.com/vivistar/synth.git
cd synth
open index.html

The interface loads with no server. AI features need a key: either supply your own OpenRouter key in the app, or run behind the serverless proxy (below).

Host it

Deploy index.html, api/ai.js, and the supporting pages (landing.html, terms.html, privacy.html, trust.html, references.html) to a host that runs the serverless function (e.g. Vercel).

Set the OPENROUTER_API_KEY environment variable so the app uses the server-side proxy with no key required from the user.


API Key & Admin Gate

  • Shared key: set OPENROUTER_API_KEY in your deployment's environment variables. The key is used server-side by api/ai.js and never reaches the browser.
  • Bring your own key: any user can paste their own OpenRouter key into the in-app key popover; those requests go directly from their browser to OpenRouter and bypass the proxy. The key is held only in browser memory.
  • Admin unlock (optional): set ADMIN_UNLOCK to a secret code and the wired OPENROUTER_API_KEY will be used only when a request presents a matching admin unlock code (entered once in the app). This reserves your key for yourself while everyone else brings their own. Leave ADMIN_UNLOCK unset to keep the shared key open to all.

See the Trust Center for full key security guidance.


Fork it

Anchorflight is open source under Apache-2.0. If you're building your own synthetic users, fork it, adapt the gates and prompts to your own stack, and self-host. Attribution requirements are satisfied by retaining the copyright notice, the NOTICE file, and a link back to this repository.


License & Attribution

Copyright 2026 Terry M. Patterson

Licensed under the Apache License, Version 2.0. See the LICENSE file for the full text, or visit https://www.apache.org/licenses/LICENSE-2.0.

Maintained by Terry M. Patterson (@vivistar)


Legal & Trust

Document Purpose
Terms of Service Permitted/prohibited use, AI disclaimer, liability
Privacy Policy No-collection posture, OpenRouter data flow, GDPR/CCPA
Trust Center Architecture, AI transparency, compliance posture, vulnerability disclosure
References Literature basis for the framework

About

Know when synthetic users help, and when they'll fool you. An interactive triage tool for weighing product decisions before you trust AI-simulated users.

Topics

Resources

Contributing

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages