One-Change Retry Loop
Turn failure into the next experiment by changing one thing at a time
- Difficulty
- Starter
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 99%
Kaku's answer to failure is a compact experimental loop: try again, but make a slight change informed by the previous result. The first attempt creates a learning curve rather than a final verdict. Start by naming the failed outcome and the most plausible source of error. Change one controllable variable, repeat the attempt under reasonably comparable conditions, and compare results. If the change moves the outcome in the right direction, retain it; if not, reverse it or test a different adjustment. Limiting each cycle to one material change improves causal learning because the result is easier to interpret. The loop produces incremental knowledge and a better next attempt, but it should not be used where retries are dangerous, irreversible, or too costly to run without review.
Origin
Extracted from The Diary of a CEO
Core principles
- 01A first attempt naturally contains mistakes
- 02Failure becomes useful when it changes the next attempt
- 03A small controlled change makes learning clearer
- 04Repeated feedback should guide the direction of later changes
How to run it
- 1
Name the Failed Outcome
Describe what happened and how it differed from the intended result. Focus on observable facts rather than treating the outcome as a judgment of personal ability.
Pro tip Write the target and actual result side by side.
Watch out Do not retry a dangerous or irreversible action automatically.
- 2
Extract One Lesson
Identify one plausible cause or constraint revealed by the attempt. Choose something specific enough to influence the next test.
Pro tip Prefer evidence from the attempt over a broad story about why you always fail.
Watch out A plausible cause is still a hypothesis until tested.
- 3
Make One Slight Change
Adjust one material variable while keeping the rest of the attempt as comparable as practical. Make the change large enough to matter but small enough to interpret.
Pro tip Record the change before starting the retry.
Watch out Changing many variables at once makes the source of improvement unclear.
- 4
Retry and Compare
Run the next attempt and compare it with the previous result. Decide whether the change moved the outcome in the intended direction.
Pro tip Use the same success measure in every cycle.
- 5
Update the Next Attempt
Keep a helpful change, reverse an unhelpful one, or select another hypothesis. Continue only while the learning and potential result justify another cycle.
Pro tip Stop when the expected value of another attempt no longer exceeds its cost or risk.
Watch out Persistence without learning is repetition, not iteration.
In the wild
Asked for the best strategy to deal with failure, Kaku says to try again because the first attempt is part of a learning curve. He recommends making a slight change, checking whether it moves in the right direction, and changing again if it does not.
→ Failure becomes feedback for a controlled next attempt rather than a stopping signal.
Illustrative example: a founder's signup page converts poorly. Instead of redesigning everything, the founder changes only the headline, reruns the test with the same audience source, and compares the signup rate. The result worsens, so the headline is reversed before the call to action is tested next.
→ Each retry produces interpretable evidence about one variable.
Common mistakes
Repeating Without Changing
Kaku's advice is not blind repetition; the learning from one attempt must alter the next.
Changing Everything at Once
A bundle of changes can improve the result while hiding which adjustment mattered.
Retrying an Unsafe Failure
Some failures require expert analysis, safeguards, or a stop decision rather than another immediate attempt.
Is it for you?
Best for
It is best for repeatable tasks where another attempt is affordable and feedback arrives quickly.
Not ideal for
It is not ideal for irreversible, unsafe, or high-stakes actions that require expert review before any retry.
From the transcript
“the best strategy to deal with failure is try it again”
“try it again, but make a slight change”
“If not, make another change”
From the episode
World-Renowned Physicist: The Truth About Aliens! UFOs Are Definitely Robotic - Michio Kaku