Featured AI experiment
WMATA Incident Summarizer
A split frontend/backend experiment for turning WMATA transit incident data into cleaner human-readable summaries.
PrototypeCode / AI Experiments
AI Experiments
Carefully scoped AI helpers, summarizers, content tools, review workflows, and experiments where the AI has to earn its keep instead of just waving jazz hands.
Featured experiment
Real data, cleaner summaries
The strongest AI projects are not magic tricks. They take messy inputs, process them clearly, and give people something easier to understand.
Active experiments
Summaries, processors, and content helpers
Current experiments focused on summarization, data cleanup, workflow support, and turning rough inputs into more useful output.
Python / AI
Weather Summary Tool
An Open-Meteo and AI summarizer experiment for turning raw weather data into readable daily summaries.
PrototypePython / AI / Web
WMATA Incident Summarizer
A practical AI summarization project built around real transit incident data, backend processing, and a separate frontend.
PrototypeContent workflow
Social Media Post Generator
A helper for turning project notes and structured inputs into cleaner social posts, descriptions, and reusable copy.
ActivePlanned experiments
Ideas for later, not distractions for today
These are AI-assisted workflows that could become genuinely useful once the surrounding projects are mature enough to justify them.
Moderation / Review
CraftForge Image Review
A future helper for flagging uploaded images that may contain protected IP, hate symbols, or questionable custom requests before production.
PlannedShop workflow
Product Listing Assistant
A practical assistant for turning real product specs, materials, print time, and pricing notes into clearer listings and posts.
PlannedPortfolio workflow
Project Writeup Assistant
A helper for turning README files, commit notes, and project decisions into human-readable build notes and portfolio posts.
PlannedAI rules
The experiment has to have a job
AI is useful when it is scoped, testable, and attached to a real workflow. Otherwise it becomes a very confident fog machine.
Rule
Useful before clever
The AI part has to reduce real work, clean up messy information, or make a workflow easier. Otherwise it is just expensive glitter.
AlwaysRule
Human review stays in the loop
For moderation, content, business, and public-facing decisions, AI should assist and flag. It should not blindly approve things that can cause trouble.
AlwaysRule
Small scope wins
The best experiments are narrow enough to test, debug, and explain without needing a whiteboard, a grant, and a nap.
Always