Article
Five archetypes, and the line between an observation and a measurement
A widely-shared way of describing how work divides up on a team, read next to what a psychometric instrument is actually able to say.
Boris Cherny, who leads Claude Code at Anthropic, described five archetypes he sees on his team in a post on 28 June 2026. He noted that the archetypes are not tied to job title, that many people span two or three of them, and that the healthy mix shifts with the product's stage.
Prototyper
Generates many ideas, most of which never ship.
Builder
Turns a prototype into a production-grade product.
Sweeper
Simplifies, unships and optimises.
Grower
Iterates a built product toward product-market fit.
Maintainer
Keeps a mature system secure, reliable and fast as it scales.
What this is, and what it is not
These archetypes are one person's observation of one team. They carry no validation of any kind: no reliability estimate, no norm sample, no evidence that they divide people cleanly or that a person belongs in one of them. They are not an Anthropic product. Equilibrium does not produce an archetype score — it does not measure these archetypes, does not assign anyone to one, and nothing in a report of ours will ever tell a person or a manager which archetype somebody is. Anything below is reasoning about the archetypes as an idea, not output the instrument produces.
Why the idea is useful anyway
What makes the archetypes interesting is exactly what makes them unmeasurable in this form: they describe a posture toward a piece of work at a moment in its life, not a fixed property of a person. The observation that most people span two or three of them, and that the right mix depends on the stage the product is at, is a statement about work, not about individuals.
An instrument works at a different level. Equilibrium measures defined constructs — things like Experimentation, Reflection and Feedback Seeking, each with a written definition and a confidence band on every score. You can read those definitions on the construct glossary. It is reasonable to read an archetype description and think about which of those constructs the described behaviour would draw on; that is a thought about the definitions, not a mapping the instrument performs. There is no calculation behind it, no score attached to it, and no grouping we assign on the basis of it.
The honest version of the idea, in measurement terms, is that a team is made up of people whose measured profiles differ, working on a product that is at some stage. Equilibrium can describe the first half of that sentence, with stated error, and only in aggregate once at least five people have responded. It cannot tell you the second half, and it cannot turn either into a label. What it can and cannot yet do is stated in full on the limitations page, and the observed evidence behind each construct is on the evidence ledger.
Measure the thing that is measurable
The free AI Adaptation Profile takes about twelve minutes and returns every score with its confidence band.
Take the free profile