The ladder stops responding before it ends, and then it stops being a ladder.

The flattening

Recent results move the pool less and less, and eventually not at all.

Within any gate, the previous cycle's outcomes move the authorised pool across a band. The band's width is not constant down the ladder. Growth spans six percentage points and recovery seven, so a good cycle is worth something material. Buffer narrows to five, floor to four. At ground-floor the span is zero — twelve percent regardless of what the last four trades did. The account's recent performance, which is the operator's most vivid evidence and the input they will most want to act on, has by that point been formally disconnected from deployment. That is not the ladder running out of resolution by accident. It is a deliberate statement that in a terminal-caution state the question of how the last cycle went is no longer the operative question.

Why responsiveness is withdrawn first

The input most likely to argue for expansion is removed before the floor is reached.

Consider what a responsive pool would do in a deep-floor state. The account has lost a substantial fraction from its peak, the mission has changed from growth to survival, and a run of four winners arrives — as it eventually will, since runs happen. A responsive ladder would authorise more capital on that evidence, at the moment the account has the least capacity to absorb being wrong about it. The recovery arithmetic is already punishing at that depth, so a re-deepening from an expansion taken on four trades is the most expensive available mistake. Flattening the band removes the mechanism rather than relying on the operator to decline the invitation, which is the system's consistent preference wherever a temptation can be engineered out instead of resisted.

The terminal state

System lock is a different kind of entry — not a smaller size, an absence of one.

Past roughly forty percent from the peak the ladder does not offer a further reduced pool. It offers no trading until review. That is a categorical break rather than a continuation, and it is worth noticing because every other state on the ladder answers the question how much with a number. This one refuses the question. The reasoning is that beyond a certain depth the recovery arithmetic has become severe enough that continued deployment at any size is a worse expected path than stopping to establish what happened, and the required review is the substance of the state — the halt is not a punishment interval to be waited out, it is a condition that ends when the system has been examined.

The asymmetry with the top

The ladder's upper end is bounded by a cap; its lower end is bounded by a stop.

The two ends of the ladder are not mirror images and the difference is instructive. At the top, growth permits the full range and the constraint is a ceiling — more evidence cannot produce more than the maximum row, but the account continues operating normally against it. At the bottom, the constraint is not a floor on size but a termination of the activity. If the ladder were symmetric there would be a minimum viable pool that the account traded at indefinitely, and the system explicitly declines to offer one. That refusal encodes a claim worth stating plainly: there is a depth past which the correct trade size is not small, it is none, and a system that always returns a number will never say so.

What it means in practice

In the bottom states, effort moves from sizing to diagnosis.

The practical consequence for an operator in the lower bands is that the deployment question has largely been answered and stops repaying attention. The pool is what the gate says, recent results will not move it much or at all, and the throttle's tier arithmetic has little room to express anything. What does repay attention is everything the diagnostic layer measures — whether expectancy is intact, whether the drawdown is damage or friction or deployment inefficiency, whether the structure is deteriorating or merely absorbing a difficult stretch. The ladder flattening is in effect the system redirecting effort: it has stopped asking the operator to make deployment decisions and started asking them to establish what went wrong.

• Band width shrinks with depth and reaches zero one state before the ladder ends.
• Responsiveness is withdrawn where a good run would be most expensive to act on.
• The terminal state answers how much with none, and ends on review rather than on time.

The key idea

A governor that always returns a number cannot express the most important answer.

Most risk frameworks scale continuously toward zero, which sounds prudent and quietly guarantees that the system never tells the operator to stop. Something is always permitted, at some size, and the operator keeps trading through the exact conditions in which trading is least likely to help. Building a discrete terminal state into the bottom of the ladder means the framework retains the ability to say no rather than merely to say less. It is the same capability the diagnostic layer gets from leaving immature cells blank: the most useful output a system can produce is sometimes a refusal, and refusals have to be designed in because they never emerge from arithmetic on their own.