Don't Stop Until You Reach the End
Distrust expert answers and drive every problem to ground-truth yourself
- Difficulty
- Advanced
- Time to result
- ~ongoing to results
- Steps
- 4
- Confidence
- 92%
A first-principles investigation discipline: the person owning execution must become the true expert on a problem and refuse to stop at the first plausible answer. 'Experts' are usually just the most-tenured person available, and they will confidently give you answers that are wrong — not maliciously, just from their own belief. You keep pushing, re-measuring, and re-instrumenting until you reach ground truth.
Origin
Ayo Omojola developed this building the Cash Card at Square — personally visiting card factories to learn laser-engraving — and applies it now at Carbon Health via relentless questioning.
Core principles
- 01An 'expert' is usually just the most-tenured person in the domain you're asking about, not the world's foremost authority
- 02People give you the answer they believe to be true — not lying, not malicious, but often just wrong
- 03You cannot stop until you reach the end of the problem
- 04The execution owner (not necessarily you) must become the expert and command all the details
- 05Optimizing something for the first time means re-measuring, re-instrumenting, and rebuilding all the queries from scratch
How to run it
- 1
Treat every expert answer as a hypothesis, not a verdict
When you ask a tenured person something, take their answer as one belief to test, not the truth. Keep asking 'what does this really mean?' and pushing until the answer holds up.
Pro tip In regulated or physics-constrained environments with many variables, this matters most — details are unavoidable.
Watch out This is expensive work — only justified when you're operating in a domain that matters.
- 2
Re-measure and re-instrument from scratch
To optimize something, don't trust prior measurements. Re-instrument it, rebuild the queries and visuals, and look at the data 15 different ways.
Pro tip Whenever two things look incongruent, go find out exactly why before moving on.
Watch out Skipping this means any success you get is 'more than likely fortune than skill.'
- 3
Ask the questions people think don't matter
Probe the small anomalies others dismiss — a null field, an unexplained value. These edge cases are where correctness breaks in complex environments.
Pro tip Frame it concretely: 'I just need us to put residual balance in that field if that's what it's for — that's all I'm asking.'
Watch out People will resist and ask why you care; expect to endure the friction to get the details right.
- 4
Put a true owner in command of the details
You can't do everything yourself, so ensure the person in the execution role becomes the expert and is accountable for reaching the end — with organizational trust to do so.
Pro tip Operationalize it as 'trust but verify' — and don't accept 'X person said so' or 'we're obligated to' without articulating exactly what breaks otherwise.
Watch out Enormous value is lost when the execution owner isn't really in command of all the details.
In the wild
The first card-vendor samples for the Cash Card looked unshippable. Omojola personally toured card-manufacturing factories, discovered laser engraving (then rare on mainstream cards), and learned the machines have thousands of setting combinations that each create a different physical texture. The team tested easily a thousand combinations of plastic, overlay, paper, envelope, and finish before shipping v1.
→ Cash Card shipped as a physically differentiated object unlike anything in market, with a team still refining physical card objects today.
A database field records why each payment was made, but a null value was used to mean 'billing the patient for a post-adjudication residual balance.' Across 120+ clinics with many humans pressing pay buttons, Omojola realized any exceptional payment reason would be silently swallowed by that null, so he pushed a resistant colleague to write 'residual balance' explicitly into the field.
→ The correction was being implemented live during the conversation, closing a data-integrity blind spot in a complex regulated environment.
Common mistakes
Accepting the first expert answer as truth
The most-tenured person will confidently give you an answer they believe, which is often just wrong. Stopping there bakes the error into your product.
Letting the execution owner skip the details
If the person accountable for execution isn't in true command of every detail, huge value leaks out and outcomes become luck rather than skill.
Is it for you?
Best for
Product leaders and execution owners tackling optimization or correctness problems in complex, regulated, or physics-constrained domains
Not ideal for
Small-stakes environments where the expensive, tedious work of driving every detail to ground truth isn't justified
From the transcript
“usually an expert is the most tenured person in the world in the domain that you're questioning”
“they'll give you an answer which is like the thing that they believe to be true they're not lying and it's not malicious and it's…”
“you just have to keep pushing till you get to an answer”
“you can still do well but it's actually more than likely Fortune than skill”
“the person who you trust in the execution role they they have to become the expert they have to be like they can't stop till…”
“I just need us to put residual balance in that field if that's what it's for that's all I'm asking”
From the episode
Frameworks for product differentiation, team building, and thinking from first principles
Ayo Omojola (Carbon Health, Cash App)