Skip to content
Judd Gurtman

Work

The Year of AI is one project at a time, shipped in public, each with a write-up of what broke. Every number below came from a measured run, and each case study says which one.

AI projects

Aug to Sept 2026Python · FastAPI · Claude API · installable web app

Coach

A full-stack AI fitness app. You talk to it like a trainer and it acts: an agent logs workouts, meals and weigh-ins through 7 tools and writes each day's session from the weights you actually lifted. Under it: a FastAPI backend, a phone app you can install, and a simulator that runs a fake user through ten days to test every change.

Coach tab: the test user reports a workout and the coach logs it with the log_workout toolWorkout tab: today's session, with bench press set at 165 lb because the last logged session was 160Journey tab: level 4, XP and daily quests computed from the logs

Read the case study → Code ↗

Tools the agent can call
7
Offline tests
37
Scorecard, run 1 → 3 of the last 5 runs
7 → 13 of 13
Cost of a 10-day test run, run 1 → run 16
$2.07 → $1.35
Sept 2026Python · Claude API · web search

Research agent

Ask it a hard question. It splits the question into smaller ones, researches them in parallel, checks every claim against the exact quote it came from, and writes a brief where every sentence cites a source.

Read the case study → Code ↗

Weakly supported claims, saved replay
66% → 54%
Sentences with a citation, tuned run
45 / 46
Invented sources, every saved run
0
Cost per question, first → tuned run
$1.68 → $1.25
Sept 2026Python · MCP · orchestration

Agent fleet

One orchestrator runs my other agents as tools through an MCP server. Anything that spends money has a cap and waits for a person to approve it, and a stuck job gets shut down so it stops costing money.

Read the case study → Code ↗

Agents connected so far
2
Search cap per question
4 × 3
Sept 2026JavaScript · Cloudflare Workers · Claude API

Judd Bot

The assistant on this site. Ask it about my work by voice or text: it answers only from a public list of facts, cites a source for every answer, takes you to any page you ask for, and says it doesn't know rather than guess. Code checks every answer before a visitor sees it.

Try it → Code ↗

Eval questions passed, run of Oct 4
100 / 100
Questions per visitor per minute
6

Data projects

Data projectPython · pandas · EIA data

Fuel price volatility

30+ years of weekly U.S. gasoline and diesel prices (26,000+ observations): how volatile they are, how regional prices drift apart, and which shocks drove the biggest swings. A rolling 12-week window, because shorter windows tracked the seasons and longer ones flattened the 2022 shock.

Code ↗

Data projectPython · APIs · SQLite

Movie ratings

A pipeline that pulls film data from the TMDb and OMDb APIs into SQLite, then compares how the two rate each genre. They disagree most on war films (TMDb rates them 1.63 points higher) and almost agree on dramas (0.09 apart).

Code ↗

For fun

Clout Royale home base: a pixel-art mansion with upgrade stations Clout Royale mission: a top-down arena fight against parody creator characters

TypeScript · Phaser 3

Clout Royale

A satirical top-down arena game about the creator economy. You play a gym influencer defending your mansion from waves of parody creators. Eliminations earn clout, your Aura meter powers your abilities, and your Cringe meter slows you down when you overdo it.

Play it in your browser (keyboard needed) ↗ Code ↗

Desktop only: keyboard and mouse.