David Strachan’s game-design essay argues the whole concept of a difficulty curve is wrong: games shouldn’t ramp up smoothly, they should be built around a difficulty saw — a series of mechanic-introduction arcs, each with its own curve. The insight came from Lemmings designer Russell Kay, who smoothed difficulty by breaking step-changes into chapters rather than difficulty settings, letting players choose and set expectations. The core problem: developers get good at their own game and lose the first-time-player mindset, so “how hard should a jump be” becomes meaningless without deliberate structure. Instead of one global ramp, group mechanics together and give each its own teach → practice → press arc — a new enemy type, ability, puzzle type, way of using a system (Mario’s triple jump), item, or rule. Smoothing means removing too-hard/too-easy spikes per mechanic, not flattening everything. The essay’s practical spine is the DO/DON’T list: optional collectables, alternate routes, dynamic difficulty adjustment (“actual magic sauce”), multiple learning channels (text, voiceover, button prompts, particles, visual cues), show-don’t-tell before giving a new weapon — but never punish bad performance with harder future attempts, never let players skip an area before learning its skill, never hide the main mechanic. Finding your skill curve empirically: death-location clusters (Mario Maker-style), level-completion deltas (80% finish L4 but 20% finish L5 → L5 is too hard or boring), structured tester questions on a spreadsheet, and watching streamed gameplay. It’s impossible to make a game for everyone — pick a target market. And the Portal 2 knowledge-shortcut trick: hide a level-skip that requires techniques learned later in the game, because if someone can trick their way to the end, the level has nothing to teach them.