‹ Drafts

John Carmack: Ideas & Engineering - Core Ideas (Compressed)

The Idea Myth

  • Ideas are vastly overvalued compared to execution
  • “An idea is worth exactly one bottle of scotch” (Heinlein)
  • The “lone inventor with brilliant idea” narrative is mostly false
  • Most “unique ideas” are either obvious to insiders, fatally flawed, or incoherent
  • A good project = hundreds/thousands of good decisions, not one eureka moment

How to Think About Ideas

  • Deliberately try to break your own ideas early - make it a game
  • “Pet ideas” are dangerous - untested ideas calcify into defended beliefs
  • Devaluing ideas psychologically helps you generate more of them
  • Be antifragile: enjoy the high of new ideas, but don’t be damaged when they fail
  • “Ride the idea high at the beginning, but it’s a puzzle to figure out how to bust it”

Where Ideas Come From

  • Ideas come from “working in the mud” on real problems
  • Creativity can be farmed - don’t wait passively for inspiration
  • Expose yourself to diverse stimuli while keeping your current problem in mind
  • Old ideas resurface in new contexts years/decades later
  • Build prototypes, not patents - physical artifacts beat theoretical claims

The 4x Rule

  • Improvements need to be ~4x to meaningfully matter
  • Marginal gains often don’t justify the complexity
  • Focus on substantial wins, not incremental optimization

On Learning & Craft

  • Math takes time - even Carmack didn’t fully understand graphics math for a decade
  • Young engineers underestimate how deeply good engineers care (bordering on obsession)
  • “Sometimes the best way to be really smart is to figure out why you don’t have to be really smart” (stupid tricks > elegant complexity)
  • Pragmatism over purity when shipping

Tactical Insights

  • Test ideas by “working from outside in” until you find what busts them
  • Iterate through multiple approaches - first attempts usually fail
  • Latency/resources can be “spent” when buying something more valuable
  • Have many ideas - need a dozen before one is great
  • More ideas than time to explore them is the ideal state