Build story

A production web platform, built and hardened with our AI-first pipeline

Mobile Lighthouse 100 performance, 100 accessibility, 100 best practices, 100 SEO. A security audit with zero open findings. Monitored around the clock, with backups we have actually restored.

Snapshot

Project
automatedstack.ai, a directory of AI and developer tools with owner listings, an admin console and a newsletter
Owner
Data Envy (our own product)
Stack
TanStack Start, React 19, Tailwind, PostgreSQL, Drizzle ORM, on our dedicated server
Status
Live

The challenge

We wanted to prove, in public, that a small team working with AI agents can ship a platform to the standard of a much larger engineering team: fast on a phone, secure by default, and run like production from day one. So we built automatedstack.ai with exactly the process we use for clients, and documented every step.

What we built

How we built it

  1. 1

    Spec first

    Each stage started as a written spec: data model, routes, server functions and acceptance checks.

  2. 2

    AI agents build

    Claude Code agents wrote the code against the spec and our engineering runbook, one feature at a time, across 11 tagged milestones.

  3. 3

    Automated gates

    Type checks, builds, a scan that keeps secrets out of the browser bundle, and performance budgets ran before anything reached review.

  4. 4

    Human merge gate

    A senior engineer reviewed and approved every change. Nothing reached production without it.

  5. 5

    Audit after launch

    We attacked our own live site: every server function was tested with and without a login, and from a foreign website.

Results

  • Measure

    Mobile performance

    Result
    100 / 100
    How it was measured
    PageSpeed Insights, mobile, median of 3 runs, November 2026
  • Measure

    Accessibility

    Result
    100 / 100
    How it was measured
    Lighthouse and PageSpeed Insights, mobile
  • Measure

    Best practices

    Result
    100 / 100
    How it was measured
    Lighthouse, mobile
  • Measure

    SEO

    Result
    100 / 100
    How it was measured
    Lighthouse, mobile
  • Measure

    Server functions audited

    Result
    21 of 21 correctly protected (11 private, all guarded; 10 intentionally public)
    How it was measured
    Live requests with and without a session
  • Measure

    Requests from a foreign site

    Result
    Blocked (403)
    How it was measured
    Live cross-origin test
  • Measure

    Backups

    Result
    Restored in full to a scratch database, zero errors
    How it was measured
    Restore drill, October 2026
  • Measure

    Email unsubscribe

    Result
    One-click unsubscribe proven end to end
    How it was measured
    Live test through the production server

What this means for your business

The same pipeline builds your platform. You get the speed of AI-assisted development with the checks of a careful engineering team: a written spec before code, human review of every change, a security audit before launch, and 24/7 monitoring after it.

Want the same process on your platform?

We would like to use analytics cookies to see which pages people actually read. Nothing is set unless you say yes, and we do not use them to identify you. How we handle your data