Case study · Started May 2025
Angry Letters
Question
Can a solo operator ship and run a full SaaS with AI support while preserving quality, governance, and operational control?
Labs
Experiments in AI-native product development. Each case study tracks question, hypothesis, system, experiment, evidence, what changed, and reusable artifacts.
01 · Angry Letters · Method and outcome
Hypothesis
A focused Product OS with explicit context boundaries and decision controls can support fast solo execution without losing accountability.
System
Cursor-centric workflow, separate repository boundaries for product code versus reusable operating assets, and clear human checkpoints for launch and operations decisions.
Experiment
Build and operate a live SaaS with payments, creator workflows, moderation operations, and AI-assisted drafting under real user constraints.
Evidence
The product is live and operating at angryletters.us, with reusable operating assets extracted into a dedicated public repository.
What changed
System separation became non-negotiable: experiment-specific operations remain product-local, while reusable operating structures are published independently for portability.
Repository, README, and INSTALL for reusable assets. Ideas behind this experiment: Thinking. Systems produced: Frameworks.
Get artifacts
Case study · Started February 2026
matthewheyman.com
Question
Can one Product OS support both a public site and a repeatable experimentation workflow without losing clarity or governance?
02 · matthewheyman.com · Method and outcome
Hypothesis
A documentation-native operating model with staged gates can keep product intent, implementation, and review aligned across multiple feature cycles.
System
Static site architecture plus lifecycle workflows, authority tiers, explicit stage gates, and canonical documentation index for durable execution memory.
Experiment
Ship multiple content, IA, and front-end changes using one operating model while preserving review gates, rollback paths, and machine-readable outputs.
Evidence
Feature artifacts, stage-gate records, and deployed static outputs demonstrate repeatability across implementation and documentation-heavy work.
What changed
Public category language shifted to Product OS while internal technical identifiers remain intact for compatibility and portability.
Repository, README, and INSTALL for reusable assets. Ideas behind this experiment: Thinking. Systems produced: Frameworks.
Get artifacts
Case study · Started September 2026
Mage Match
Question
Can a tight vs-CPU loop prove a puzzle-battle is fun enough to rematch before building networked play?
03 · Mage Match · Method and outcome
Hypothesis
A focused vs-CPU Play loop can prove rematch value before spending time on rooms, accounts, or ranked play.
System
TypeScript game-core independent of the renderer, Vite client, and a product shell that keeps Campaign and friend rooms behind later gates.
Experiment
Ship a playable browser prototype with vs-CPU Play as the public first cut. First to two. No account required.
Evidence
The product is live at magematch-web.onrender.com. Core loop, roster, and presentation shipped. Campaign and multiplayer remain gated.
What changed
Do not skip to networked play. Prove the build-break-mana-spell loop in a solo match first, then open human PvP.
Product repository only. No portable OS pack yet. Ideas behind this experiment: Thinking.
Get artifacts