When Sam and I committed to building 30 games, we knew we needed a process that was fast but not sloppy. Three days per game felt right — long enough to build something real, short enough that we couldn't overthink it. Here's exactly how it works, using real examples from games that shipped (and several that didn't).
Day 1: Concept and Paper Prototype (Sam)
Sam starts every game with a single constraint: the mechanic must be explainable in one sentence with no conjunctions. "Match falling colors before they hit the ground" — that's Color Rush. "Drop blocks with perfect timing or they get sliced" — that's Tower Stacker. If a concept requires "and" or "but" in its one-sentence description, it's too complex for a browser game and gets killed immediately.
He then draws the game on paper. Not a sketch — a playable paper prototype. For Block Breaker, he drew a grid of squares on a notepad and used a coin as the ball, sliding it across the page with his finger while I called out angles. We played 20 rounds on paper before I wrote a single line of code. The paper prototype revealed three things the spec didn't: the paddle needed to be wider than we thought, the ball speed curve needed an S-shape (slow start, steep middle, plateau at high speed), and the power-up drop rate felt right at 22%.
About 60% of concepts die on Day 1. Sam killed a game called "Rhythm Stack" where blocks fell to music because the paper prototype revealed that without actual audio synchronization — which we couldn't do in a 12KB file — the mechanic felt random rather than rhythmic. Good kill. The graveyard of Day 1 concepts is the reason our shipped games are tight.
Day 2: Core Mechanic (Tom)
I code the core mechanic in one sitting, usually 6-8 hours. The rule: by midnight on Day 2, the game must be playable from start to game-over with no features beyond the single mechanic. No scoring, no particles, no audio, no high scores. Just the loop.
For Gravity Flip, the Day 2 build was literally a rectangle that fell toward whichever surface gravity was pointing at, with a jump button. That's it. No obstacles, no score, no particles. I played it for 20 minutes and the mechanic felt right — the disorientation of gravity flipping every 5 seconds was genuinely fun even with no game around it. If the core loop isn't fun in its naked form, adding features won't fix it.
The hardest Day 2 I've had was Word Blitz. The word validation system needed a dictionary, input filtering, and timer logic — three interdependent systems that all had to work before the game was playable. It took 14 hours and I finished at 3 AM. But when it worked and I typed "MOUNTAIN" from 7 random letters for the first time, the satisfaction was instant. Worth the extra hours.
Day 3: Polish, Test, Ship
Morning: scoring system, high score persistence, audio. Afternoon: particle effects, screen shake, visual polish. Evening: Sam plays 20 rounds and files bugs, I fix them, we ship.
The audio is always last because it takes 20 lines of Web Audio API code and makes the game feel twice as polished. A simple oscillator tone on collision, a descending sawtooth on death, a celebratory arpeggio on victory — that's the entire audio budget for every game, and it's always worth the 200 bytes it costs.
What gets cut: levels, progression systems, unlockable content, story. Every game we've shipped is a single-screen, single-mechanic, infinite-loop experience. We tried adding a level system to Snake Arena once — 5 progressively harder mazes. Sam tested it and said the loading screen between levels killed the retry loop. We cut it. The game is better as one endless arena.
What We Learned From 45 Attempts
We've started 45 game concepts. 15 shipped, 30 died at Day 1 or Day 2. The common thread among the 30 that died: they sounded fun on paper but the core mechanic wasn't satisfying when stripped bare. The 15 that shipped all share one trait: you can explain them to someone in one sentence and they immediately want to try it.
If you're building browser games, kill more concepts. Ship fewer, better titles. The best game we ever made — Tower Stacker — is also the simplest. One mechanic, one screen, infinite replayability. That's the formula.