I build products,
and I sweat the details.

Treating pricing like a product is the right idea. But the back end, the month spend runs away from someone, is the part almost no one builds.

As money rails go invisible, they feel exactly like a bank while carrying none of a bank's protections. The recourse disappears along with the word crypto.

The approval/fraud/cost balance is scored entirely from the merchant's side of the counter. The false positive is a real person the system decided to read as a threat.

Underwriting moved from a human's gut to a rigid model to a data-rich re-underwriting. At every step, the question that matters is who falls outside the box, not who gets in.

Incumbents do not lose to fintechs because of COBOL. They lose because they are governed to prevent failure in a market that rewards reducing the cost of it. The bottleneck is the org chart, not the stack.

Enormous AML spend, a tiny catch rate, and the people it hurts most are the ones who cannot prove they exist at all. Compliance is necessary. The way we do it now is not working.

As money moves itself, banking dissolves into the invisible back end of other people's products. Frictionless access is fine until the bad month. No one is building the embedded path to forbearance.

A swipe is two journeys at different speeds: authorization in seconds, settlement in days, fees extracted in between. The one mental model every fintech PM should be able to draw from memory.

First-principles banking: if you started from scratch today, what would you build? The embedded-finance thesis is right. But banking everywhere means default everywhere, and the futurists skip that part.

Composability and the five flaws of centralized finance. The most extreme front door yet: permissionless, global, always on. It also deletes the back end entirely, and that is not a solved problem.