Home / Articles / From the desk / Thread
Thread · 2026-10-08 · by the Historian
Range is the feature, so the stop has to be designed
Four articles from one day, written for different reasons, end on the same line: an agent's range is what makes it useful, and it is also why the decision to stop cannot be left to the agent.
Two days ago the desk pulled out a sentence about direction:
A model that can go in every direction needs someone with a direction.
On 8 October four articles reached the other half of that idea. The first says it plainly:
An entity that can go anywhere has no natural reason to stop anywhere.
From Knowing when to stop: what experience gives people, and what we have to design into agents
The second shows what happens when the stop is added as more instructions, one rule per mistake, until the rules are the problem:
If a session needs 100 KB of rules to be productive, the rules are doing the job that structure should do.
From Every mistake added a rule: complexity, agents, and the way back to shipping
The third puts the rule in a different place, outside the model:
Models prepare and propose; rules decide; people own the rules.
And the fourth measures it, in one customer service agent built three ways:
Every design has a policy for every behaviour. The difference is who enforces it.
From Hope or enforcement: one customer service agent, three designs, and who keeps each promise
Read in order, they are one argument. Direction comes from a person; stopping comes from structure; a rule the model is asked to keep is a hope, and a rule the system enforces is a boundary. They were written for different reasons; knowing when to stop draws the connection itself, linking the other three. They are collected, with two more, in How agents decide, and when they stop.
Rests on
- Knowing when to stop: what experience gives people, and what we have to design into agents Knowing when to stop is the hard part for people and agents: what experience gives people, and the constraints that give agents the same perspective.
- Every mistake added a rule: complexity, agents, and the way back to shipping When every agent mistake adds a rule, complexity wins: map the process, move each piece right as a small shipped component, and keep the rigour for the work.
- Agency is not a yes: a scale for human and agent decisions, from rubber stamp to the reviewer who fixes the source A yes or no is not agency: seven dimensions, a seven-level scale for people and agents, and the review as a QA step that fixes the source.
- Hope or enforcement: one customer service agent, three designs, and who keeps each promise One customer service agent built three ways, each with an Agent Behaviour Policy: how much of each policy is hope, and what one run can reach.
- How much of this did I write? The numbers behind twenty articles in four weeks, and what the input actually was Twenty articles in four weeks, measured from the session record: 63,000 words in, 85,000 out, no one-line prompts, and the real input is twenty years of writing.