How I think
Find the constraint. Find the gap. Test the fix that earns its cost.
Three problems, one method. A pond in Durham, a ward in Kampala, and a growth-stage company. Short breakdowns of how each one went.
Case 1 · The log that ate lures
Why do my cousins catch more fish than I do?
Constraint. Same pond, same water, same hours. The only variable was the lure and how it was worked.
Data. A submerged log snagged lures for years. I recovered dozens and sorted them by shape, color, weight, and hook position. The ones that caught fish shared a profile. The ones that only caught fishermen were the flashy ones.
Test. Designed lures in CAD around the winning profile, printed them, and fished them side by side with the originals. Kept the two that outperformed. Changed one variable at a time. Cheap to print, so cheap to be wrong.
Result. More fish, fewer snags, and a habit I did not have a name for yet.
Case 2 · A water chamber in Kampala
Why is a proven therapy failing in the ward?
Constraint. bCPAP works. Nurses cover up to 20 neonates each. The machine needs water, and the chamber runs dry without anyone seeing it.
Gap. No alert existed. The failure was not clinical or technical. It was attention, and attention was the scarcest resource on the ward.
Test. 80 concepts, narrowed by morphological chart and Pugh matrix to a capacitive sensor with a light and a buzzer, in a wooden box that can be built in Kampala for under $15. Dropped it 35 times from 1.5 meters to make sure it survives a ward.
Result. A documented device handed to Makerere faculty for local replication. The constraint was the design brief.
Case 3 · Does it scale, or does it just work?
How do you read a growth-stage company?
Constraint. Past the earliest risk, before mature metrics. A good pitch and a good business look similar from the outside.
Gap. Most decks describe the system at its current size. The diligence question is what breaks at 10x: unit economics, regulatory path, supplier concentration, the founder’s own bandwidth.
Test. Map the market first so the company is positioned against something real. Then ask the engineering question of each claim. Where is the load path, and what fails first under it?
Result. A memo that says what has to be true, and what we would have to see to believe it. The same shape as a test plan.