One person. 16 production apps. One factory that builds them all.
I'm working through a backlog of 250+ product ideas — evaluating, building, and shipping to beta users as fast as possible. The trick isn't speed. It's the system: shared infrastructure, reusable packages, and compounding knowledge that makes each new product trivially easy to launch.
This GitHub is that system in the open.
Most people's GitHub is a graveyard of weekend projects. Mine is a production factory — I can go from idea to deployed product with real users in a single session.
| Repo | What It Proves |
|---|---|
| kits | 12 backend packages, 447 tests — extracted from 16 production apps. Build once, ship everywhere. |
| teek | Persona simulation engine. Not "act like X" prompts — structured cognitive profiles that model how someone thinks. 8-section schema, source-grounded, predictive. |
| itchy | Byte-level 16MB language model. The only byte-level submission in 622+ OpenAI Parameter Golf entries. Built the right size instead of shrinking the wrong one. |
| ai-factory | CLI that solves the AI coding agent amnesia problem. Persistent knowledge store (FTS5 SQLite) + session injection = compounding returns on every hour invested. |
| polyfeeds | 106 live data feeds mapped to prediction market categories. Scraping, APIs, and pipelines across 10 data streams — running in production. |
Every repo here follows the same philosophy:
- Build something that works — ship it, use it, find the pain
- Extract what compounds — the piece that transfers gets pulled into a shared package
- Make the next thing trivially easy — because the hard part is already solved
This is how one person ships 16 production apps across iOS, web, and watchOS. Not by working harder — by making every session build on the last one.
- 250+ ideas in a structured pipeline — evaluated, scored, built or killed
- 16 production apps shipped (iOS, web, watchOS)
- 12 shared packages with 447 tests
- 106 live data feeds in production
- 20 playbooks documenting solved problems
| Repo | What It Is |
|---|---|
| roundtable | AI persona engine — 16 cognitive profiles, adversarial idea evaluation, 51% kill rate on 199 ideas |
| agentspawn | Psychology-grounded framework for designing first interactions between AI agents and humans |
| findthemoney | How NYC's $115B budget actually works — tracing public dollars without double-counting |
Stack: TypeScript, Swift, Python, Node.js, Next.js, SQLite, Fastify, SwiftUI
Contact: pauljump@me.com · LinkedIn