Skip to content
Stupid Experiments / lab

the person running this lab

Samarth Saraswat

AI engineer at Tazapay, building fraud and anti-money-laundering detection for cross-border payments. Before that, three-plus years at Microsoft on LLM copilots and agents. IIT Guwahati. Occasionally holds a microphone and tells strangers jokes.

This lab is where I stop saying “someone should test that” and test it myself.

Lab record

experiments
0
confirmed
0
busted
0
still weird
0
inconclusive
0

clean slate.

How stupid experiments work

  1. 1questionCapture something interesting.
  2. 2clarifyWhat am I actually trying to understand?
  3. 3researchExisting evidence and competing views.
  4. 4hypothesisA working theory I could be wrong about.
  5. 5experimentDesign it and actually run it.
  6. 6logDecisions, failures, surprises.
  7. 7findingsWhat happened, stated plainly.
  8. 8reflectWhat this changes in my thinking.
  9. 9next?The next question becomes an experiment.

House rules

  • Curiosity should lead to action.
  • Test ideas instead of just discussing them.
  • A failed experiment is fine if I learn something from it.
  • The process counts as much as the result.
  • Writing is how I think, not content.
  • Show evidence, not manufactured expertise.