TL;DR - The highest-leverage engineering decisions aren't technical. They're about the product. Find the most frequent user flow, pour ~80% of the effort there, use AI to prototype visually before writing code, and run agents in the background so your time multiplies. Nobody cares if the code is clean; they care whether it moves the business.
Find the most important flow - it's the most frequent one
The most important flow usually comes from the product manager, and it's almost always the thing done most often in the system. For one product it's resetting a password; for another it's creating users. The whole application should be designed and oriented around those core steps.
Prioritize by business value = frequency of use
Deprioritization is simple - how often is it actually used? If the core business value is "create users fast," then ~80% of the effort belongs there, and everything else gets what's left. That's how you avoid burning a day polishing something nobody touches.
Decide what NOT to build by seeing the whole picture first
Before building, I iterate through the entire picture and try to see it.
- I use AI to generate a screenshot / mockup so I can enhance my creativity without touching code.
- I look at the mocked-up popup and decide "this validates the phone number with X library," and so on.
- Reviewing a cheap visual first tells me what's worth building before I've spent an hour building the wrong thing.
Multiply your time - run agents in the background
- The day before, I prepare exactly what needs to be done.
- In the morning I run it, or run it autonomously while I work on something else.
Letting well-scoped agents work in the background on prepared tasks is how I get far more out of the same hours.
The throughline
Engineering effort should follow business leverage, not code aesthetics.
- Understand the product.
- Find the highest-frequency flow.
- Prototype it visually before coding.
- Put the effort where the value is.
- Let background agents multiply the rest.