Two educational products, built from scratch: problem, thesis, research, build plan, and the working app.

MathClash and SplitHomework are educational web apps I built to take a product idea all the way to running software. MathClash is a 1v1 math duel that wraps a real difficulty ladder in low-stakes, near-peer competition. SplitHomework is a split-screen study tool that gates a YouTube feed behind completed homework — built for the phone-shaped attention span, and especially for students with ADHD. Different problems, but the same approach behind both: a narrow claim, the research to back it, a realistic plan to productize it, and a designed-and-coded app that demonstrates the loop.

Each one came with a full set of product artifacts, not just code. For both apps there's a product thesis (one deliberately narrow, falsifiable claim, plus the bets that would disprove it), a research brief (every quantitative claim cited to a real source — what couldn't be verified is flagged, not invented), and a productization build plan (a strategic fork that picks a customer, a gap analysis against the existing tools, a phased roadmap, team and budget, the top risks, and a build-vs-partner call). Then the actual app — UI and front-end code — running live.

The honest framing: these are demos. No users, no efficacy data, scope kept deliberately narrow — and the build plans and research briefs say exactly that. What they demonstrate is the full arc done by one person: noticing a problem, turning it into a thesis, grounding it in real research, planning the productization without hand-waving the cost, and shipping working software — design, code, and the writing in between.

MathClash is a 1v1 math duel. You pick a topic, you're matched against another player, and you race each other through a stream of progressively harder questions — difficulty 1 to 10, points scaling with difficulty, a worked solution shown whenever you miss one (with a Skip button), and a one-minute clock per question. After enough rounds it shows an ACT-style "where you'd stand" percentile: your accuracy maps to a raw score out of 60, then a 1–36 scaled score, then a percentile. The opponent is simulated on the player's own machine but presented as a live player.

The thesis it's built around: near-peer, low-stakes, anonymous, repeated 1v1 competition wrapped around a real difficulty ladder is one of the few motivational structures with consistent evidence behind it for confident, high-achieving students — and no consumer product is squarely aimed at the capable kid whose school won't track them and whose family isn't plugged into the talent-search world. The build covers roughly 2,000 generated, correct-by-construction questions with worked solutions across ten topics, the percentile pipeline, and a mobile-friendly UI.

SplitHomework is a split-screen study tool. Homework on the left — runtime-generated math problems with worked solutions — and a YouTube panel on the right that stays locked until you earn time. A correct answer banks one minute, a wrong one thirty seconds, capped at ten; the clock ticks down while you watch and the panel re-locks at zero. The feed is curated educational playlists across sixteen categories (the sciences, history, languages, and more), reached from a few quick-pick chips plus a custom dropdown.

The thesis it's built around: short-form video has trained a generation's reward system to expect fast, frequent payoffs, and for students with ADHD the mismatch with slow, low-feedback schoolwork is the wall they hit nightly. Rather than fight that wiring with abstinence, SplitHomework channels it — the instant-reward feed is gated behind completed schoolwork, which is the contingency-management / Premack ("first, then") structure with the strongest evidence behind it for this population. It's framed as lowering the activation energy to start and re-engage with homework — the real ADHD pain point — not as "training multitasking," which the literature does not support. The research brief grounds each claim in a real source and is explicit about what isn't proven.

The full 0→1 arc, done by one person. Problem → thesis → research → build plan → designed-and-coded app → deployed. Each of these is a real running thing with the strategy work in front of it, not a deck about a product that doesn't exist. The same loop I'd run inside a company — just compressed, and visible end to end.

Discipline in the writing, not just the code. The research briefs trace every quantitative claim to a real URL and flag the things that couldn't be verified rather than inventing them. The theses are written as one narrow, falsifiable claim with the bets that would disprove it. The build plans pick a customer, name the gap honestly, and don't hand-wave the cost or the risk — including, for SplitHomework, the uncomfortable fact that the in-app lock is bypassable and a real version needs a browser-level block, which the plan says outright.

And an honest sense of scope. These are demos: no users, no efficacy data, deliberately narrow. The point is the method — spotting a problem, sharpening it, grounding it, planning it without illusions, and shipping working software — and the range it covers: product thinking, research, strategy, UI design, and front-end code, in one person's hands.