24+ essays on product, AI, and marketplaces.

Goals tell the team what everyone already agreed on. Non-goals record the decisions that will otherwise vanish.

If you can read your roadmap from left to right and predict every meeting before it happens, you don't have a roadmap. You have a calendar.

A spec describes what to build. A decision describes what you considered, what you rejected, and why. One ages well. The other doesn't.

MVPs aren't defined by what's in them. They're defined by what removing one more thing would break.

If you cannot describe the user problem in five messages a real user would write back to you about, you don't have a problem statement. You have a hypothesis dressed up as one.

Every product team writes success metrics. Almost none writes the metric that would tell them to stop. That asymmetry is why dead products keep getting funded.

RICE is a calculator that produces fake precision. Most real prioritization is two-by-two and a quiet, honest conversation.

What the user sees when the AI is wrong is more important than what they see when it's right. Spec the fallback first.

Curated test sets always perform better than production data. If you're not testing on production-shaped inputs, your benchmark is a press release.

Most AI features fail in production not because the model was wrong — but because nobody asked the right questions at scoping.

Models drift, providers throttle, prompts break. If your feature has no plan for what to do when the AI is unavailable, you don't have a production feature.

The cases where adding AI actively makes the product worse. Recognize them before you spec.

Every AI feature should be tested against one question: does this remove a step the user is already taking? If no, kill it.

Build evals when wrong outputs cost money, customers, or trust. Skip them when you can fix a bad output in five minutes.

Most teams that 'need' a vector database actually need keyword search and a small JSON file.

The phrase is overused. The actual work has three distinct phases, each with different skills, and PMs who confuse them fail.

Minimum, viable, product. All three words mean something. Most teams ship versions that miss one of them.

Product-market fit is real. It doesn't look like a metric improving. It looks like users acting different.

Marketplaces obsess over checkout. They lose to competitors that figured out the trust layer first.

Throwing marketing dollars at an empty marketplace is the most common mistake in two-sided platforms. The early problem is supply quality, not demand volume.

A marketplace with thin liquidity that scales gets thinner liquidity, not thicker. Fix liquidity first.

Your supply is not your customer. But it's the one that leaves first.

Every successful marketplace started by manufacturing one side. The 'organic' supply story is rewritten after the fact.

Two-week sprints work for consumer products. They don't work for enterprise. The math is structural, not philosophical.