Skip to content

Speed Comes From Killing Bad Ideas Faster

Chronicles entry — originally published externally and preserved here as a historical milestone.

Context

  • Project: crushr
  • Series: Chronicles
  • Status: Historical snapshot
  • Original venue: LinkedIn

Speed Comes From Killing Bad Ideas Faster

March 18, 2026

Four days ago I had a compression/archive format experiment with 0% recoverability.

Today I have a settled format direction and a clear understanding of what works, what doesn’t, and why.

Article content

What changed wasn’t a breakthrough idea. It was the loop.

I stopped trying to design the right format and instead treated every change like a test:

  • make a structural change
  • break it on purpose
  • measure what survives
  • keep it or kill it

I used AI to speed up the build/test cycle—generating variants, wiring things up faster, keeping iterations tight—but the decisions were always based on the results.

Most ideas didn’t survive contact with damage:

  • centralized metadata looked clean but failed quickly
  • adding more metadata usually made things worse
  • structural recovery and usable recovery turned out to be different problems

Once enough bad ideas were eliminated, the remaining design space got small enough that the format effectively stabilized.

That’s where things stand now: not done, but no longer guessing.

Next step is building a proper benchmark harness to see how it behaves against real data:

  • compression ratios
  • performance
  • behavior under stress

The takeaway for me:

If you can make your tests honest and comparable, and you can iterate fast enough to trust the results, you don’t need a perfect idea—you just need to remove the wrong ones quickly.