The four muscles
Something ambiguous arrives. You turn it into intent, you get it done through more than your own hands, you decide what was worth doing at all, and you move people who do not report to you. Everyone runs this loop. What changes with seniority is the size of the thing that was ambiguous when it arrived.
Each muscle below has the same four parts: what it is in plain words, the one test that separates it from activity, how it scales as the work gets bigger, and the ways people sincerely believe they have it when they do not. That last part matters most. Nearly everyone fails on the fakes before they fail on the definition.
Phase one
Turn ambiguity into intent that AI and people can execute against.
Synthesis is not summarising. Summarising compresses what already exists and points it upwards. Synthesis takes a fog of half-stated needs, contradictions and unspoken constraints, and produces something specific enough that somebody else, or a model, can act on it and be right. It points downwards and outwards, at the people who have to do the thing.
It is the phase most people skip, because the fog is uncomfortable and starting work feels like progress. Every downstream failure that gets described as a communication problem is usually a Synthesis failure with a later date on it.
Something arrived vague and left specific enough that others, or a model, could act on it without coming back to you. Execution proceeded, and what came out matched. The person who judges the match is the executor, not you.
| Size | What you are making specific |
|---|---|
| Ticket | One piece of work, clear enough that nobody asks a follow-up question. |
| Deliverable | A whole thing, where you wrote the spec that nobody handed you. |
| Outcome | What the business actually needs, translated into work that will get there. |
| Whole area | A domain nobody had defined, now legible enough that a team can be pointed at it. |
| What the job even is | The problem itself was wrong, and you are the one who reframed it. |
Phase two
Multiply yourself through AI, tools, and people.
Leverage is the difference between output that stops when you stop and output that keeps going. It has nothing to do with how hard you work, and quite a lot to do with whether the thing you built is still running in your absence.
This is the muscle where AI has changed the floor most violently. Work that used to demonstrate leverage now demonstrates almost nothing, because a model does it in a sentence. The test has moved accordingly: it is no longer about what you produced, it is about who else can now produce it.
Never "I built a tool". Somebody else ran it, unaided, with you not in the room.
| Size | Whose capacity went up |
|---|---|
| Your hands | You do the same work faster. Real, but it stops with you. |
| Your week | Something you used to do repeatedly now happens without you touching it. |
| Your team | People who are not you use the thing you left behind, unaided. |
| The function's capacity | The way the whole function works changed, and would not revert if you left. |
Phase three
Decide what's worth doing, and whether it's right.
Judgment is the only muscle whose evidence has two timestamps. You have to have decided before the answer was knowable, and the reasoning has to still hold up after the result came in. Either half alone proves nothing. Deciding early with bad reasoning is a coin toss. Explaining well after the fact is narration.
This is also the muscle that most rewards writing things down, for an unglamorous reason: memory quietly edits your reasoning to match the outcome, and it does it sincerely.
You decided before the answer was knowable, you wrote down why, and the reasoning is still legible after the result came in. Two timestamps, not one.
| Size | What you were deciding |
|---|---|
| Within a task | Which approach, given a constraint nobody had spelled out. |
| Within a deliverable | What to leave out, and defending the cut. |
| Within an outcome | What to build at all, when several options were defensible. |
| Across a system | Which of several real problems does not get solved this quarter. |
| Across quarters | Which bets are worth making before anyone can tell you they were. |
Phase four
Move outcomes through people you don't control.
Influence is the strictest of the four, because the evidence is not yours to produce. It sits in somebody else's record. Their roadmap changed, their budget moved, their queue got reordered, and they would say so themselves without you standing there.
Authority is the thing most often mistaken for it. If someone did it because you or your manager asked, that is compliance, and compliance is a perfectly good way to get things done. It just does not place you anywhere.
Somebody you don't control did something different, and they'd say so themselves. The evidence lives in the other party's record, not in yours.
| Size | Who moved |
|---|---|
| One person | A peer changed what they were going to do. |
| Another team | A different function's plan, budget or queue changed. |
| A standard | People you never spoke to apply something because it became the way things are done. |
| Outside the company | A client, partner or market participant moved, and a colleague watched it happen. |
How to read your own
You do not have a Synthesis number. You have wins, and each win came out of one phase of the loop more than the others. Tag them honestly and a pattern appears, usually not the one you expected.