💡 In simple words: When people build things with the help of AI, job titles start to matter less than talents. A researcher who created a famous AI coding tool says there are five big talents: inventing new ideas, building them for real, cleaning up and simplifying, helping the product grow, and keeping it running safely. WikiDeal would like to test this way of working: one person brings lots of ideas, and a team with the four other talents would help turn the best ones into something everyone can use.

🎯 In 20 seconds (expert summary): Methodology page, proposed as an initial hypothesis. Boris Cherny (Anthropic, creator of Claude Code) describes five work archetypes that emerge as AI blurs the traditional product roles: Prototyper, Builder, Sweeper, Grower and Maintainer; the archetype, not the job title, describes what a person actually does. The WikiDeal R&D programme intends to adopt this framework as an initial hypothesis, to be tested during Prototype 1: open calls would invite people matching these five profiles to participate in the definition, the accompaniment and the implementation of the open calls and of their results. Open calls combined with the five profiles are the envisaged lever to move from idea to prototype to deployment, in short to scale. In the startup phase, Theo Bondolfi plays the Prototyper role (ideas and content production); the other roles are intended to be progressively aggregated into the operational team.

AI-supported deployment

Status: under construction. This page presents an initial hypothesis, proposed as a basis for discussion, to be tested during Prototype 1.

Context: when AI blurs job titles

In teams that build products with strong AI support, the traditional split between engineering, product management, design and data science appears to blur: increasingly, everyone directs the same AI agents. Boris Cherny, creator of the Claude Code tool at Anthropic (the company behind Claude), described this evolution in a widely shared post: in his words, engineering, product, design and data science "melt into a new kind of role". What then distinguishes people is no longer their function but the phase of the product lifecycle in which they thrive, and he proposes five archetypes to name those phases: Prototyper, Builder, Sweeper, Grower and Maintainer.

A key point of Cherny's framework is that these archetypes are not tied to job titles: at Anthropic, some designers, engineers, product managers and data scientists each match any of the five patterns. The archetype describes what a person actually does, not what their badge says.

The framework also has an older lineage. One published analysis traces it back to Robert X. Cringely's 1992 split of tech companies into "Commandos, Infantry and Police", to Simon Wardley's "Pioneers, Settlers and Town Planners" (of Wardley map fame), and to Peter Thiel's "zero to one" versus "one to n" distinction (Zero to One). According to that analysis, the novelty would not be the five boxes themselves, but the fact that the lifecycle phase becomes the primary axis for describing roles, instead of a secondary one.

What recent studies say about AI acceleration

Recent empirical studies give a nuanced picture of how much AI actually accelerates a project. A controlled experiment with GitHub Copilot found that developers assisted by an AI pair programmer completed a coding task 55.8 percent faster than a control group (Peng et al., 2023). By contrast, a 2025 randomized controlled trial by the research organization METR, with 16 experienced open-source developers working on mature projects they knew well, found that allowing early-2025 AI tools increased completion time by 19 percent, against the developers' own expectations. Anthropic's 2026 analysis of about 400,000 Claude Code sessions observed that people make most of the planning decisions (what to do) while the AI makes most of the execution decisions (how to do it), and that on coding tasks people from non-engineering occupations succeed at nearly the same rate as software engineers.

Read together, these studies suggest that AI support can strongly boost a project but does not do so automatically: how the human roles are organized around the AI appears to matter. This is one of the reasons why WikiDeal intends to test an explicit organization of roles, described below, rather than assuming that AI alone would accelerate the work.

The five profiles

For each archetype, the first sentence follows Cherny's original definition; the second describes, prudently, the role this profile could play in a project like WikiDeal.

1. Prototyper. Comes up with brand new ideas and churns out many of them, most of which do not ship. In WikiDeal, this profile would generate concepts and content: portal drafts, policy ideas, contract-model hypotheses and innovation pages, prototyping applied to socio-economic design as much as to code. Many drafts would remain hypotheses, and that would be normal.

2. Builder. Quickly turns a prototype or an idea into a production-grade product and infrastructure. In WikiDeal, Builders would turn mockups and concepts into a production-grade platform: the wiki, the funding tools and the marketplace applications, moving from a minimum viable product to real software deployment.

3. Sweeper. Cleans up the user interface, simplifies the code and the system, unships features and optimizes performance. In WikiDeal, Sweepers would simplify pages, structures and code, remove what is superfluous, and keep technical debt under control.

4. Grower. Takes a product that has been built and iterates on it to improve product-market fit. In WikiDeal, Growers would iterate toward community and market fit: adoption of the portals, contract models and marketplaces by real users and User Groups.

5. Maintainer. Owns a mature system to make it secure, reliable, fast and efficient as it scales. In WikiDeal, Maintainers would look after the reliability, security and performance of the wiki and of the marketplaces at scale (software maintenance, scalability, in the spirit of DevOps).

In this reading grid, the same person can combine several archetypes, and none of them is reserved to a specific job title.

The methodology: five profiles plus open calls

This approach is the emerging methodology that the WikiDeal R&D programme intends to adopt, as an initial hypothesis. It would be tested during Prototype 1 (see Prototype 1 deliverables).

Concretely:

the open calls would include calls toward these five figures; there is a real intention to try to fill these roles, at least in the form of a pilot experiment;
the way the pilot experiments are intended to work is to invite these figures to participate in the definition, the accompaniment and the implementation of the open calls and of the results that follow from them;
the methodology of the open calls, combined with the five figures, is the envisaged lever to move from the idea to the prototype and to deployment: in short, to scale.

Whether these roles can actually be filled remains to be seen; that is precisely what the Prototype 1 pilot experiments aim at testing.

Roles and credits

The startup of the project can be read through the same grid:

for the startup phase, Theo Bondolfi plays the role of bringing the ideas and producing a large volume of content, which corresponds to the Prototyper archetype;
the other figures and roles are intended to be progressively aggregated into the operational team, notably through the open calls described above.

Roles and origins of the concepts are credited on the Licensing and Credits page.

Sources and references

Framework sources:

Boris Cherny, original post on X: https://x.com/bcherny/status/2071379474277613732 (also on Threads: https://www.threads.com/@boris_cherny/post/DaJgVFVj2PB/)
Video summary (French): "Le créateur de Claude Code vient de dire que les anciens métiers sont obsolètes", Henri, ExplorIA, YouTube: https://www.youtube.com/shorts/pNHnjGhCMW4
Analysis and lineage of the framework: "The Archetype Under the Title", paddo.dev: https://paddo.dev/blog/the-archetype-under-the-title/
Analysis of the strengths and traps of each archetype: "Five Archetypes of Workers in the AI Age", braindetox.kr: https://braindetox.kr/en/posts/claude_code_work_archetypes_2026.html

Scientific references on AI-supported acceleration:

Peng et al. (2023), "The Impact of AI on Developer Productivity: Evidence from GitHub Copilot", controlled experiment, arXiv:2302.06590: https://arxiv.org/abs/2302.06590
METR (2025), "Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity", randomized controlled trial, arXiv:2507.09089: https://arxiv.org/abs/2507.09089 (summary: https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/)
Anthropic (2026), "How Claude Code is used in practice", privacy-preserving analysis of about 400,000 agentic coding sessions: https://www.anthropic.com/research/claude-code-expertise
Anthropic Economic Index, ongoing research initiative on the effects of AI on work and the economy: https://www.anthropic.com/research/the-anthropic-economic-index

Gov/en/Portal:R&D/Innovations:AI Supported Deployment