Engineering, written the way we'd say it out loud.
Real trade-offs, specific stack choices, and honest takes on when AI is — and isn't — the right tool. No listicles.
Building Rulebook Trader: Where the Rules Decide and AI Just Reviews
A paper-trading side project built for fun, for testing strategies, and for trying tools. Deterministic rules make every call, a cheap AI reviewer can only veto, and every decision gets logged. The pattern travels well beyond trading.
Where Small Businesses Should (and Shouldn't) Use AI
AI pays off where you have a pile of messy text or a decision repeated hundreds of times — and it wastes money everywhere else. A three-question test, with real examples from a dealer-network build.
Building SwingSync: Comparing Two Golf Swings That Don't Run at the Same Speed
An iPhone app for studying two golf swings side by side — synced by phase, not by the clock, with a local pose-estimation backend that refuses to guess timing markers more often than it offers one.
Building Hearth Index: Forecasting Home Prices for Every ZIP Code, Not Just the Metro
A one-day feasibility spike that forecasts home values for 894 metros and 26,268 ZIP codes, backtested against nine years of real outcomes. The two-part model, the ablation discipline, and the bugs it caught along the way.
Building Alpaca Explorer: Asking the Market a Question Instead of Reading a Dashboard
A chat interface over live Alpaca market data, with Claude running a read-only tool-use loop instead of a wall of tickers. The architecture, the "unusual options activity" heuristic, and what I'd build next.
Building Cellar Index: Teaching a Model Which Wines Are About to Change Price
Ten years of government PDF price lists turned into three production models — and the ablation discipline that rejected most of the features I tried. Plus what's still left to tune.
Building MoodLens: a mood journal that reads between the lines
How I built an iPhone app that reads the emotion in what you write or say, keeps your data private, and costs almost nothing to run.
What a 4.16-Star Average Doesn't Tell You
Building a review-intelligence prototype for a multi-line hospitality business: sentiment scoring, zero-shot aspect classification, and what 1,531 real reviews revealed once broken out by business line.
Building PhotoNester: Teaching an iPhone to Organize a Million Memories
An app that sorts your entire camera roll into Travel, Pets, Food, and more — without a single photo ever leaving your phone. Here's how it actually works, and the bug that almost shipped a broken app.
When to Build a Custom Model vs. Use an API
OpenAI's API is the right call more often than most technical founders want to admit. Here's the actual decision tree, not the sales pitch.
Have a problem that isn't in a post yet?
Tell us what you're building — we'll tell you honestly what we think.