Decision OS / Experiments
When the full answer is unknowable, design a small test that reality can grade.
Turn stuck decisions into time-boxed tests with a question, action, signal, and decision rule.
Case notes
Experiments makes the decision process visible before the outcome arrives.
Some decisions cannot be solved in the head. You do not know how a role feels, how a routine fits, how customers respond, how a conversation lands, or what you prefer until reality answers back.
An experiment is not procrastination if it has a decision rule. It is procrastination when the test never ends or the signal is ignored.
01
Write the question first.
The experiment must know what it is trying to learn.
02
Choose a signal before acting.
Otherwise you will reinterpret results to protect your preference.
03
Time-box the test.
An endless experiment is just another delay.
Common problems and experiments
Make the next process move small enough to test this week.
I need certainty before trying.
Experiment
Choose the smallest safe action that could teach something this week.
What to watch
Some certainty only appears after contact.
My experiment is too vague.
Experiment
Write question, action, time box, signal, and decision rule.
What to watch
A real test needs a finish line.
I ignore the result.
Experiment
Decide in advance what each signal means.
What to watch
Precommitment protects learning.
Prompt to try
Keep one decision sentence visible.
What small test would produce information I cannot get by thinking longer?
7-day protocol
The seven-day decision experiment
- 01 Write the decision uncertainty.
- 02 Turn it into one question.
- 03 Choose one small safe action.
- 04 Define the signal you will observe.
- 05 Write what you will do if the signal is positive, mixed, or negative.
- 06 Run the test within seven days.
- 07 Update the decision, not the story.
Decision checklist
Mark the process, not the outcome.
Source notes
Implementation intentions
Specific if-then plans help convert experiment signals into action.
Open source →Uncertainty
Judgment under uncertainty benefits from checks that reduce reliance on untested impressions.
Open source →Education-only scope
Experiments should be safe, bounded, and not used where urgent expert or emergency action is required.