Triage was invented for war.
Napoleonic field hospitals had more wounded than hands. Surgeons sorted casualties into three categories. Those who would die without help. Those who would survive with it. Those beyond saving. What looked like brutal arithmetic was the only way to save the most people, given finite hands.
Solo founders running on AI face a smaller version of the same problem. More problems than hands. The arithmetic is similar. The instinct should be too.
What AI actually shifted
Most writing about the solo AI founder is about output. One person matching what used to take many. That is the productivity story. It is also the wrong story. Productivity gains compound, then plateau, then everyone has the same tools and the gain is competed away.
The interesting part of the model is not how much one person can produce. It is what one person has to leave undone.
That is a triage problem, not a productivity problem.
Before AI was useful, the question was whether something could get done. Now the question is what gets left undone. Both are problems. They are different problems.
The old bottleneck was code output. The new bottleneck is attention. AI writes migrations, tests, emails, support replies. What AI does not do is decide the order. The order is the entire job.
The old bottleneck was code output. The new bottleneck is attention.
The three stamps
Triage is older than software. The three categories transfer cleanly.
Red is what is on fire now. A bug blocking payments. An abuse pattern tripling server cost. A new lead with a time-bound need. The hand goes here.
Yellow is what can wait. A larger refactor. A new feature idea. An onboarding flow to optimize. The ticket gets written. The state gets documented. Then it gets parked.
The critical part of yellow is that parked is not forgotten. Parked means the context survives the wait. This is where most solo founders fail. They park inside their heads, not inside systems. Three weeks later the context is gone, but the problem is red again.
Black is what is not getting touched today. A lead that has been sitting for three days. A feature request from a low-tier customer. A bug that only hits edge cases. It gets stamped black. Not because it does not matter. Because no hand is free, and nothing else is bleeding worse.
The hard part is not recognizing red. The hard part is stamping yellow and black as such, and not flipping them at midnight because a yellow problem suddenly feels red.
Triage is not prioritization
Prioritization is soft. Triage is hard.
A priority list says A before B before C. It allows that after A, B might still get touched if there are twenty minutes left. Triage does not allow that. Triage says anything below red goes untouched today. Even if energy is left over. Energy goes into the next red item, not into yellow.
Prioritization lets things drift up. Triage lets them drift down. Down is what saves you.
The reason is structural. In the field hospital, energy is not the constraint. Attention is. Working at half-focus on a yellow injury misses the red one tipping over. The same applies to a company with multiple frontiers.
What actually helps
Externalized context. Every yellow item has to leave the head and live in a system that can be trusted tomorrow. The tool is interchangeable. What matters is that picking up later does not mean remembering. It means reading it back.
Structured asymmetry. Not every platform has to be maintained the same way. A listing platform with a quarterly update cycle is not the same as a SaaS in daily operation. The permission to run platforms differently is not weakness. It is a precondition. Treat them all the same and the load is the same everywhere.
AI as triage helper, not triage replacement. AI is good for execution. Not for the decision. What next stays with the founder. How to ship this clean and fast goes to the model. The split matters. Letting AI decide what is red gets reliably the wrong reds.
Accepted fires. Some things have to burn. A detail page that does not render perfectly for a few days. A report that is one day late. An onboarding that stays clunky until there is air. A solo operator who tries to keep everything green is in burnout in two months and out of business in four. Accepted burning is risk management. Not sloppiness.
What triage is not
Triage is not a way of life. Anyone running permanently in triage mode builds nothing. Only repairs. That does not last.
Triage is a transition mode. It buys time for structures to emerge that need less of it. Better monitoring. Cleaner processes. Documentation that allows pickup.
Treating triage as endpoint is a different problem than treating it as phase. From the outside they look the same. From inside they are not.
The test is whether the share of red incidents per quarter stays flat, rises, or drops. Flat or rising means something is wrong in the architecture. Dropping means the right things are getting built.
What is left
The marketable image of the solo AI founder is a person who replaces fifteen employees and sits in a café while agents make money. That image exists. Rarely. In very asymmetrically constructed businesses.
The realistic image is a person with three stamps at the breakfast table. One who does not fix everything during the day, but decides what is allowed to stay broken.
AI did not solve the solo founder problem. It created a new one that demands sharper operators.
Anyone who accepts that can run several businesses. Anyone who does not, burns. Hotter than the fires they meant to put out.
.svg)
