A dream is a plan with feelings attached. Every other plan you hold gets reviewed on a schedule, and this one never does, which is strange, because the resource it spends is the only one that does not replenish.
The unit is attempts, and you get few
Getting genuinely good at something takes years. Finding out where its ceiling sits takes a few more, because a ceiling is invisible until you are close to it — from underneath, a capped domain and an uncapped one look identical, and both feel like progress.
So a serious attempt at a domain runs close to a decade, and a working life holds a small, countable number of those. Not dozens. Few enough that the number should sit slightly uncomfortably when you write it down.
That single reframe changes what every other decision means:
- Sunk cost inverts. “I have already given this ten years” is not an argument for continuing. Those ten years are spent under either choice. The only live question is what the next ten buy, and continuing does not recover the investment — it doubles it.
- Staying is a decision, not a default. Every year you do not audit is a year you chose the current answer without checking whether it was still the answer.
- Delay is the expensive option. The teardown gets more costly the longer you wait, because the structure standing on top of it gets larger, and because the attempts remaining get fewer.
None of this is an argument for restlessness. It is an argument for scheduling the review, so the decision gets made on evidence rather than by exhaustion.
What “pays off” has to mean
If you are going to choose what pays off, the phrase needs to survive contact with something other than money, or it collapses into chasing the highest fee, which is its own dead end.
The honest version is: what does one unit of your effort produce, and for how many people, and for how long? Three questions, and money is only one input to the answer.
Two of them do most of the work.
What does it cost to realise one idea here? In construction, an idea costs materials, labour, permission and thousands of dollars before anyone can see whether it was any good. In code it costs attention. That difference decides how many attempts you get inside one lifetime, and the number of attempts decides nearly everything downstream of it. I have set the two disciplines side by side on exactly this, and the comparison is more useful than the story.
Does work done once serve one person or many? Delivered work either adds or compounds. A rendering serves one client once. A system, a script or a written procedure serves everyone who touches it for as long as it survives. That is the same shift that moved the bottleneck across three generations of my family from land to access to judgment, and it is the property worth optimising a career around.
Notice what is missing from both: enjoyment. That is deliberate. Enjoyment is a bad instrument here because it tracks competence, and a teardown temporarily removes competence. You will not enjoy the first stretch of anything. Judge the destination on its ceiling; the enjoyment catches up later.
The skill is not the dream, it is the vehicle
The reason people cannot run this audit honestly is that the skill has become the answer to who they are, and you cannot price something you are.
So separate them, because the market already has. Nobody buys a skill. They buy an outcome, and the skill is one way of delivering it, priced at whatever the cheapest adequate delivery costs. That price is set by how many people can do it, what the tooling makes possible, and how much of the value the buyer can perceive. All three move, none of them ask you, and no amount of being excellent moves them back.
There is a clean test, and it takes a second.
If the same outcome could be delivered another way at a fraction of the cost, would you switch? If the honest answer is no, you are committed to the mechanism rather than to the outcome — and the mechanism is the part that gets repriced.
I failed that test for a while. The vehicle I had chosen for making things had fees that did not move, revisions with no end condition, hardware replaced on someone else’s schedule and software that billed in the quiet months. What I actually wanted was to make things I could picture. That want was completely portable. The tool it had attached itself to was not.
Tearing down is not starting over
The reason people stay past the point of sense is that demolition looks like losing everything. It looks like going from senior in one place to beginner in another, in public.
It is much less than that, because the foundation does not come down with the building.
What transfers is the part that took longest to learn: reasoning under constraints, knowing what finished means before starting, and the standard you hold yourself to. I have set the two disciplines side by side on exactly this, in both directions, so I will not repeat the list. What does not transfer is the surface: tools, vocabulary, the names of things. That layer is real, it is the smaller half, and it is the half that looks like the whole job from outside. It is also the fastest to rebuild, because you already know what you are looking for.
So the person who tears down and rebuilds is a beginner in a narrow band and a senior everywhere underneath it. My geometry did not stop being useful when I stopped drawing buildings; it runs underneath infrastructure work every week.
Distinguishing a ceiling from a bad quarter
This argument is dangerous if it is used to justify walking away from anything difficult, so it needs a hard edge on it.
A ceiling can be written down. The fee is capped by what the outcome is worth to the buyer. The cost base rises on a schedule you do not control. The number of people who can do this is growing. Those are structural statements, they are checkable, and they stay true next quarter.
A bad period cannot be written down that way. It produces sentences about how you feel, not about how the economics work. If the case for leaving does not survive being written as three factual claims about the market, it is a difficult stretch, and the answer to a difficult stretch is not a new domain.
Run the audit when you are not in pain, because a review conducted during a bad month produces the answer the bad month wants.
The part that is genuinely courage
Most of this is arithmetic: price the ceiling, price the teardown, compare. The courage is needed for something narrower — being visibly worse at your job for a while, in front of people who knew you when you were good.
That is a real cost, and it is the only part of the equation that is temporary.
Holding a skill loosely is not being casual about it. I was as serious about the craft as anyone, and I am about the current one. The looseness is only about the label: refusing to let this is what I am answer a question that should be answered by this is what I am trying to produce.
That is also what keeps readiness portable. Skills that exist only inside one context reset when the context changes; skills held as capabilities move with you, and they are the ones still there when something worth doing opens up. The same holds for how work gets defined now that agents produce so much of the output — the durable half is stating what done means and building the check for it, which belongs to no domain in particular.
Years from now the answer to the audit is usually fine. The expensive years are the ones where it was not, and nobody asked.
Be excellent at the vehicle, and never let it become the destination. When did you last write your ceiling down as three facts about the market?