Learn In Fact, Not In Theory
Use reversible real-world trials before locking in predictions
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 6
- Confidence
- 95%
Sivers prefers discovering needs through use rather than predicting them from convention. Preserve optionality first: begin with the smallest usable setup, keep pipes, wires, and other infrastructure visible, and avoid finishes that make later changes sacred or expensive. Then inhabit the design and notice what is genuinely missing. Add a desk, sink, room, or service only when experience demonstrates the need and location. The idea draws on Stewart Brand's claim that buildings are predictions and predictions are wrong, and Sivers extends it to life experiments. The framework accepts bounded mistakes because even a failed trial converts uncertainty into fact. Its guardrail is reversibility: experimentation is most appropriate when changes can be undone and safety or legal requirements do not demand a complete upfront design.
Origin
Extracted from The Tim Ferriss Show through Derek Sivers's experimental home and his application of Stewart Brand's How Buildings Learn.
Core principles
- 01Predictions about future needs are often wrong
- 02Flexible infrastructure keeps mistakes cheap to correct
- 03Direct experience reveals requirements that planning cannot
- 04A stable personal base can increase tolerance for bounded failure
How to run it
- 1
Name the prediction
Identify the requirement you are tempted to assume, such as where a sink or desk should go.
- 2
Preserve reversibility
Keep the structure easy to alter and avoid hiding infrastructure that future changes will need.
Pro tip Prefer exposed, accessible systems over sealed finishes.
Watch out Meet safety, structural, and legal requirements before experimenting.
- 3
Start minimally
Create the smallest setup that permits real use without filling it with conventional defaults.
Pro tip Bring only what you know you need at the start.
- 4
Use before deciding
Live or work in the setup long enough for repeated friction and desire to reveal actual needs.
- 5
Add from evidence
Install or change only what the lived trial shows is necessary, placing it where use indicates.
- 6
Keep learning cheap
Retain enough flexibility that the next mistaken prediction can also be corrected without rebuilding everything.
In the wild
Sivers is building simple rectangular sleeping and waking spaces near Wellington. He plans to begin with a mattress, an outdoor induction hob, an outhouse, and exposed infrastructure, then discover through use whether and where he truly needs a bathroom, kitchen, desk, or other additions.
→ The house becomes an adaptable experiment instead of a fixed bundle of default assumptions.
Common mistakes
Making the prototype sacred
An award-like finish can make a place too precious to alter, defeating the learning mechanism.
Experimenting through one-way doors
The method becomes dangerous when failure cannot be reversed or bounded.
Is it for you?
Best for
People designing spaces, systems, or life changes where their true future needs are unclear.
Not ideal for
Irreversible, safety-critical, or legally constrained decisions that require complete planning before use.
From the transcript
“all buildings are predictions and all predictions are wrong so therefore the less predict you can make your building the better”
“finding out in fact instead of just in theory”
“I'm willing to mess up happily because I will know that then I found out in fact”
From the episode
#777: Derek Sivers, Philosopher-Entrepreneur — The Greatest Year of His Life
Derek Sivers