Prototyper, Builder, Sweeper, Grower, Maintainer: The Roles Replacing Job Titles in AI Startups

Job titles may be the next thing AI coding agents make obsolete. In late June 2026, Boris Cherny — the creator and head of Anthropic's Claude Code — argued that as "engineering, product, design, DS, etc. melt into a new kind of role," the useful unit of work is no longer the department you belong to but the mode you're operating in. Looking at his own team, he mapped five archetypes that cut across every function. It's one of the sharpest public descriptions yet of how AI is reorganizing not just how software gets built, but who builds it.

The five archetypes

Cherny's framing is that these are patterns of work, not job descriptions — and crucially, none of them is tied to a function. On the Claude Code team, some designers behave like archetype 1, some like 3; the same spread holds for engineers, PMs, and data scientists.

Why AI agents dissolve the old boundaries

The reason this is happening now is mechanical. When writing code was the bottleneck, specialization made sense: you needed dedicated engineers because turning intent into working software was slow and hard. AI coding agents collapse that cost. When a designer can prompt a working prototype, a PM can ship a fix, and an engineer can spin up a landing page in an afternoon, the walls between "who does what" stop paying for themselves. The scarce skill becomes judgment about what to build and when — which is exactly what the five archetypes describe.

Anthropic has structural evidence of this: as observers noted amplifying Cherny's post, the company gives most of its people the same title — Member of Technical Staff — rather than splitting them into engineering, product, and design tracks. The PM codes, the designer codes, and the org chart flattens around capability instead of function.

Timeline: How AI Agents Moved From Chatbots to Business Automation

What it means for how startups organize

For a startup, the practical implication is that you may be over-hiring for titles and under-hiring for modes. A team of five specialists who each only do their function is less flexible than a smaller team of generalists who can each move between prototyping, building, sweeping, growing, and maintaining as the product demands. It also reframes seniority: the most valuable people aren't the ones who own a function, but the ones who can recognize which mode a problem needs and switch into it.

This doesn't mean specialists disappear — a Maintainer running a system at scale still needs deep expertise. But the default org design is shifting from "hire a function" to "cover the five modes," and small teams increasingly cover several modes per person.

FAQ

Is the "builder role" an actual job title?

No. Cherny's point is the opposite — these are archetypes of work that cut across engineering, product, design, and data science, not new titles. At Anthropic most people share one title, Member of Technical Staff.

Do the five archetypes replace specialists?

Not entirely. They describe modes of work; a person can occupy different archetypes at different times, and some roles (like Maintainer at scale) still reward deep specialization. The shift is toward people who can move between modes.

Why is this happening now?

AI coding agents drastically lower the cost of turning an idea into working software, which erodes the specialization that existed because coding was slow and expensive. When anyone can build, judgment about what to build becomes the differentiator.

Bottom line

Cherny's five archetypes — prototyper, builder, sweeper, grower, maintainer — are a useful lens precisely because they ignore job titles. As AI agents make building cheap, the org chart is reorganizing around modes of work rather than functions, and the people who thrive will be the ones fluent in switching between them. Whether or not the exact five stick, the direction is clear: in AI-native teams, what you do this week matters more than what your title says.

Sources and further reading