DeepDream creator Alex Mordvintsev weighed in on a fresh round of AI coding experiments where the same prompt produced wildly different runtimes depending on how much reasoning effort the model was allowed, prompting his claim that heavier reasoning often leads to shorter final code outputs.
My whole life experience screams that the correlation between effort and lines of code must be negative
Fable - same prompt, different reasoning efforts - 12 mins vs nearly 2 hours between 'low' and 'max'
The Fable tool demo showed one prompt finishing in twelve minutes at low effort versus nearly two hours at maximum effort, leaving builders to weigh whether the extra compute delivers proportionally better results.
Mordvintsev’s observation about reasoning effort and code length remains an open assertion without the specific prompt details or output comparisons attached, so the pattern still needs broader verification across tasks.
No Digg Deeper questions have been answered for this story yet.
@zzznah If it's a finite/closed problem, for sure... I think this one in particular is more open-ended though.
My whole life experience screams that the correlation between effort and lines of code must be negative
@zzznah "If I had more time, I would have written a shorter letter."
@zzznah Not always? An elegant solution might be a lot of time simple, but robust code requires handling edge cases, formatting errors, outputting traces, etc.
@zzznah Same, maybe this also explains why their writing is bad/verbose/feels “low entropy” to an experienced human reader
@zzznah only when adjusted for number of features.
@zzznah for the same effect. but it is not indicated what was delivered.
@zzznah Writing code is easy. Writing less code that does the same thing... that's where the effort actually goes.