Keep It Neat — an ecosystem of ad-free, no-tracking apps
Founder & sole developer. A brand, three live apps, and a shared design system — all vanilla, no trackers, no ads, all open source on GitHub.
Developer · Writer · Madison, WI
Ashlie Crooks
Full-Stack Developer · AI Workflow Engineer
I build production web apps in Rails and React, and I engineer AI workflows that make building them faster and more reliable. Six years shipping software, three of them customer-facing at a consultancy, paired with a daily practice of designing specs, validation pipelines, and multi-agent systems in Claude Code. And I write about how I do it.
Selected work
CALL Survey Platform Redesign
AI-assisted rewrite of a survey platform used by 2,000+ schools — tens of thousands of lines in three weeks. Figma AI to secure buy-in, Claude to convert components and replace legacy CoffeeScript. A force multiplier, not a substitute for architectural judgment.
Playlist Curation Engine
An AI playlist builder with a validation pipeline that took track hallucinations from ~20% to 0%. Three-tier curation, verification against Last.fm and Tidal APIs with confidence tiers. Wrote the whole tool myself, using Claude as a Python tutor.
Multi-Agent Coaching System
Two coordinated AI instances on a 200+ line behavioral spec — structured handoffs, domain context loading, cross-session persistence via a shared vault. I diagnosed and fixed real failure patterns (context drift, silent data loss) with targeted guardrails.
Spec-Driven AI Development
A repeatable pipeline for building software with AI, not just prompting it: brainstorm → versioned spec → bite-sized plan → a fresh agent per task with two-stage review. Prompts are treated like code — typed inputs, eval gates, every rule a guardrail against a mistake the model made once.
Writing
Treating Prompts Like CodeJun 2026
About
I'm a full-stack developer in Madison, WI, with six years shipping production software — Rails, React, Node, PostgreSQL. Three of those years were customer-facing at Ten Forward Consulting, where I was a technical contact for multiple clients: joining discovery calls, scoping solutions, and shipping production code across several client codebases at once.
What sets my work apart now is that I don't just use AI — I engineer with it. I write behavioral specifications precise enough that a fresh model produces correct results on first load; design layered context architectures; and build validation pipelines that catch failures before they ship. On one project I found an AI was hallucinating ~20% of its outputs; I built a verification layer that drove that to zero. I'm honest about where these tools break, because knowing that is the whole job.
I have a BA in Creative Writing, which turns out to be an asset for both halves of this work — clear technical communication for async teams, and the precision that good prompts and good specs actually require. I'm open to full-time or contract roles, remote or Madison-based.
Contact