hraness

saved

What my dad taught me about AI coding in the 90s

by Mike van RossumAskmike.orgpublished

gist

Mike van Rossum uses a 1990s blindfold-chess lesson from his father as a model for working behind Claude Code. Strong players track relationships rather than a photographic board. AI coding is the opposite: you can stop attending to interfaces and still get output, and that is vibecoding. The same attention, pattern recognition, and mental model of how code expresses an abstract goal let a programmer stay in control without reading or writing every line. Unlike chess, that skill will be hard to rank because models can satisfy any metric.

ideas

  • Blindfold skill is relational, not photographic. Strong players track attacks, structures, and what changed; knowledge reconstructs the position instead of holding a picture of the board.
  • Vibecoding is skipping the board. Little prompting and no reading of generated code produces software that will not last.
  • AI coding is the opposite of blindfold chess. You do not have to attend every turn, remember the important pieces, or track interfaces and APIs.
  • The same skills still win. A deep mental model of how code weaves an abstract goal lets someone thrive behind Claude Code without reading or writing it.
  • The skill will be unrankable. Chess has ratings; models will satisfy any quantifiable programming metric you throw at them.

quotes

you are vibecoding and the resulting software is not going to last very long

Mike van Rossum, defining vibecoding by skipped prompting and unread generated code.

the skills that define a strong blindfold chess player are the same as those of a programmer who can thrive behind a Claude Code terminal

Mike van Rossum, stating the governing analogy.

being able to have a deep mental model (in your head) of how code weaves together

Mike van Rossum, naming the transferable pre-LLM skill.

programming effectively this way will become impossible to measure

Mike van Rossum, on why the skill cannot be ranked like chess.