Find where it fits
I know what the organization actually does, day to day. That is what tells you which work AI can take on, which it can't, and which it would quietly make worse.
About
Why I know where AI fits.
Since 2018 I've been Manager of Accreditation at Social Current (formerly the Council on Accreditation), running accreditation for the Public Agency, Opioid Treatment and Child & Youth Development business lines — supervising a team of coordinators, owning the budget for two of those lines, and helping update the child welfare standards themselves. Before that, six years in foster care at MercyFirst and Forestdale as a caseworker, supervisor and advocacy liaison. I still review case records part-time at Graham Windham to stay close to practice.
Plenty of people can build a tool. Far fewer have sat in the PQI meeting, run the case record review and walked an agency through accreditation — and can therefore tell the difference between software that takes work off the table and one more system nobody updates. That combination is hard to hire for.
And not only on side projects.
At Social Current I spotted where AI could fit in my own department, made the case for
it, and had the proposal approved. Same judgment, applied inside an organization
rather than to a weekend build — and it had to survive the people who would live with
the result.
Claude Code and Codex closed the last gap. Knowing where it fits used to end at a recommendation; now it ends at something working. And reviewing work against a standard was already the day job — reading an agent's diff is the same instinct pointed at a different artifact.
Which is why a public standards library leads this page, ahead of the board game companions. Same method throughout; the difference on that one is that I already knew the standards by heart.
Read the full résumé
What I offer
From "where would this even help?" to staff actually using it.
I'm not the developer. I'm the person who knows the work well enough to say what should be built — and who stays with it until people are using it.
Brief the developers
I translate between the people who know the program and the people who write the code. They build it; I make sure what gets built matches the work it is meant to carry.
Make it effective
Shipped is not the same as working. Once it exists I tune it against real cases until the output is good enough for staff to trust — the same review discipline I already apply to case records.
Teach the staff
Adoption is where most of this fails. I train the people who will use it every day, in their language, so it becomes how the work gets done instead of a login nobody opens.
Method
How these actually get built
I don't advise on tools I haven't used in anger. The agents write most of the code; that only works because everything around the writing is tight — and reviewing the diff is the real job.
Spec before code
Decide the product, the data model and what "done" means first. A vague prompt produces a vague codebase, fast.
Agents implement
Codex and Claude Code do the typing — features, refactors, test suites — working against the repo's own conventions.
Review is the work
Every diff gets read. Typecheck, unit tests, lint and Playwright runs gate the merge, not vibes.
Ship to the edge
wrangler deploy and it's live worldwide. No VPS, no container, no
3am pager.
Live on the web
Projects
A public standards library, board game companions and tabletop reference tools. All of them online right now.
Social Current Standards
standards.social-current.org
A public, browsable home for Social Current's accreditation standards — turning a dense reference document into something an organization can navigate.
Built with Codex + Claude Code
Moonrakers Command
www.moonrakersapp.org
Companion and analytics platform for the Moonrakers board game: build your crew, track a live game turn by turn, then dig into charts, Elo ratings, player comparisons and playstyle profiles afterwards.
Built with Codex + Claude Code
TM Stats
www.tm-stats.com
Statistics for Terraforming Mars — log your games, then see how corporations, maps and table sizes actually shake out across a real play history.
Built with Codex + Claude Code
D&D 4e Parser
dnd4eparse.org
Takes D&D 4th Edition character files and turns them into a clean, readable character sheet you can actually use at the table.
Built with Codex + Claude Code
AD&D Bestiary
adnd-bestiary.fochizzy.workers.dev
A fast, searchable monster reference for Advanced Dungeons & Dragons — built for looking something up mid-session without losing the thread.
Built with Codex + Claude Code
Lotus Blossom District Atlas
lotus-blossom-district-atlas.fochizzy.workers.dev
A navigable atlas for a tabletop campaign setting — the places, the people and the connections between them, kept in one place the whole table can browse.
Built with Codex + Claude Code
Google Play
Android apps
All three are in closed testing on Google Play — access available on request.
Built with the publishers in the loop. The makers of both Moonrakers
and Valeria are aware of these apps and the companion site, and the work has been
shared with them directly.
Moonrakers Command
com.fochizzy.moonrakers
The Moonrakers companion on your phone — crew selection, turn order and live game tracking at the table, syncing with the same analytics that power the web app.
Built with Codex + Claude Code
Valeria Randomizer
com.fochizzy87.valeriarandom
Randomized setups for Valeria — deal a fresh, valid combination of cards and components so no two games open the same way.
Built with Codex + Claude Code
Valeria Score
com.fochizzy87.valeriascore
Scorepad for Valeria — runs the end-game tally across every scoring category so the maths doesn't fall to whoever is quickest with a pen.
Built with Codex + Claude Code
Stack
What it all runs on
Cloudflare Workers end to end. Static assets and rendered apps sit on the same edge network, deploy from the same CLI, and cost roughly nothing at rest.
Cloudflare Workers
Workers Static Assets
Wrangler
Custom domains & edge TLS
Next.js
OpenNext
React
React Native + Expo
TypeScript
Supabase
Vitest
Playwright
Recharts
Open to work
Looking for a role helping a non-profit integrate AI.
If your organization is trying to work out where AI fits — or has already built something and cannot get anyone to use it — that is the work I want.