Taking action
What is the difference between a reversible experiment and a permanent decision?
A reversible experiment is a bounded action designed to answer a specific question before you make a larger commitment. It has an exit path, limits on cost and impact, an observable result, and a review point. A permanent decision commits resources, obligations, access, or other people in ways that may be impossible or unacceptably costly to restore. The key difference is not whether the action feels small. It is whether you can stop, recover an acceptable prior state, and use what you learned before the stakes expand.
Separate reversibility from size
Amazon's decision framework distinguishes consequential, irreversible or nearly irreversible choices from changeable, reversible choices. Its practical point is that one decision process should not be applied to everything: a reversible choice can use a lighter process because course correction remains available.
That distinction is useful beyond a company, but it is not a complete personal decision rule. A small action can still expose private information, affect another person, create an obligation, or be costly to undo. A large-looking trial can sometimes remain reversible when its boundaries and recovery path are explicit.
Know what makes an action an experiment
A reversible action becomes an experiment when it is designed to reduce a named uncertainty. Stanford d.school describes a prototype as a way to investigate an idea and explore what could come next; its no-build approach emphasizes using a simple context rather than elaborate technology.
Formal evaluation can demand much more rigor than a personal trial. UK Cabinet Office guidance on randomized controlled trials says to decide the desired outcome in advance and compare an intervention with what would have happened without it. A personal experiment is not an RCT, but the narrower lesson transfers: define what you are trying to learn before acting, and compare the result with a clear baseline or alternative.
Run the BOUND check
BOUND is a Forever Free Compass framework, not a validated assessment. Write one concrete answer for each part before calling a choice reversible.
- Boundaries: cap the time, money, visibility, access, and other exposure before the test starts.
- Outcome to learn: name one uncertainty and the observable result that would reduce it.
- Undo path: describe how you would stop and what an acceptable restored state would look like, including the real cost of getting there.
- Needed consent: identify anyone affected and obtain agreement where the test touches their time, information, property, responsibilities, or choices.
- Decision date: set when you will review the evidence and choose to stop, adapt, repeat, or make a larger commitment.
Compare the two decision types
Use this comparison as a planning aid rather than a declaration that a choice is objectively safe or permanent.
| Question | Reversible experiment | Permanent or hard-to-reverse decision |
|---|---|---|
| Purpose | Learn before expanding commitment | Choose and accept the larger commitment |
| Scope | Capped in advance | Effects may extend beyond the initial action |
| Exit | Stop condition and recovery path are explicit | Restoration may be impossible or unacceptably costly |
| Evidence | Observable result and review date are defined | Evidence is gathered before commitment; later review may not undo it |
| Other people | Impact is bounded and consent is addressed | Rights, duties, or reliance may continue after the choice |
Do not experiment with stakes that cannot be contained
Do not use the language of experimentation to bypass consent, professional obligations, safety safeguards, legal requirements, or the consequences of exposing another person. If the action concerns health, legal rights, substantial finances, immediate safety, confidential information, or duties of care, get appropriately qualified guidance before acting.
If the BOUND check reveals no credible undo path, high recovery costs, open-ended exposure, or important effects on people who have not agreed, treat the choice as hard to reverse. Slow down, verify the critical facts, involve affected people, and make the commitment deliberately rather than trying to force it into an experiment.
Original contribution
The Forever Free Compass take
Forever Free Compass uses the BOUND check as an original planning framework: Boundaries, Outcome to learn, Undo path, Needed consent, Decision date. It distinguishes a genuine experiment from a commitment that has merely been given a temporary label. BOUND does not prove that an action is safe or reversible. It asks you to specify the maximum exposure, the evidence you need, what an acceptable recovery would require, who must agree before the test begins, and when you will decide to stop, adapt, repeat, or commit. If one answer is missing, redesign the test or treat the choice as a higher-stakes decision.
Sources
- 2016 Letter to Shareholders — Amazon. The distinction between reversible two-way-door decisions and irreversible or nearly irreversible one-way-door decisions, plus different decision processes for each.
- The No-Build Hack: Create a scrappy prototype to test an idea and learn something new — Stanford d.school. Prototypes as tools for investigating ideas, exploring what may come next, and learning with a deliberately simple setup.
- Test, Learn, Adapt: Developing Public Policy with Randomised Controlled Trials — UK Cabinet Office and Behavioural Insights Team. Predefining the desired outcome and using a comparison to determine whether an intervention is working.
This is educational reflection, not medical or mental-health care. If distress is persistent, severe, or involves immediate safety, contact a qualified professional or local emergency service.