跪拜 Guibai
← All articles
Frontend · Backend · Artificial Intelligence

Your Job Title Is a Liability: The 5 Roles That Replace 'Full-Stack'

By 仿生狮子 ·
Read original on juejin.cn ↗ Google Translate ↗ Alt translation

When Meituan merges frontend and backend teams and Shopify's AI agent ships production PRs, the question is no longer whether AI will change team structure but what structure replaces the old one. Cherny's five-archetype model gives engineering leads a concrete alternative to the full-stack generalist — one that maps people to product stages rather than technology layers.

Summary

Chinese tech giants Meituan and Ant Group are already merging frontend and backend roles and retraining testers as full-stack engineers. Shopify's AI agent River now co-authors one in every eight merged PRs. These moves are not isolated experiments; they signal a structural shift in how software teams organize work.

Boris Cherny, who built Claude Code at Anthropic, proposes that the old waterfall division of labor — product, design, frontend, backend, testing — is collapsing into five role archetypes defined by how someone creates value, not which technology they wield. Prototypers generate ideas fast and discard most of them. Builders land things in production. Sweepers delete dead code and simplify systems. Growers run the hypothesis-validation loop to find product-market fit. Maintainers guard mature systems with deep domain expertise.

A team's mix of these archetypes shifts with the product's lifecycle: exploration demands Prototypers and Builders, growth pulls in Growers, and maturity leans on Maintainers. The Sweeper is needed from day one. The framework separates the posture of value creation from the technology stack used to deliver it — a distinction that makes the 'full-stack engineer' label look increasingly obsolete.

Takeaways
Meituan's Keemart R&D team has merged frontend and backend into a single unit, with frontend engineers receiving backend training before the change took effect.
Ant Group's online banking unit is transitioning all testing roles to R&D positions with a six-month window; testers become full-stack engineers afterward.
Shopify's AI agent River lives in Slack, reads code, runs tests, submits PRs, and publicly challenges proposals; it co-authors one of every eight merged PRs company-wide.
Boris Cherny's five archetypes are Prototyper (rapid idea-to-tangible-demo), Builder (lands things in production), Sweeper (deletes code, simplifies, un-ships features), Grower (runs A/B tests and chases PMF), and Maintainer (guards mature systems with deep expertise).
At Anthropic, most staff share a single title — Member of Technical Staff — because the same person can exhibit different archetypes depending on the project and phase.
Team composition should follow product stage: exploration needs Prototypers, Builders, and Sweepers; growth adds Growers; maturity leans on Sweepers, Growers, and Maintainers.
Sweepers are essential from day one; Maintainers join as the product matures.
The archetype model answers 'how you create value,' while the full-stack model answers 'what technology stack you use to deliver value' — they are independent dimensions.
Job titles like 'Web Frontend' are losing descriptive power because they measure neither a person's real organizational influence nor their dynamically changing role.
Conclusions

Cherny's framework quietly demotes the full-stack engineer from a destination to a legacy label. If coding is a solved problem, then mastery of multiple technology layers is table stakes, not a differentiator — what matters is how someone moves a product through its lifecycle.

The Sweeper archetype is the most counterintuitive and possibly the most undervalued. Deleting 250,000 lines of code in a year, as Cherny did, is a form of engineering leverage that few job descriptions ever reward, yet it directly reduces maintenance burden and cognitive load across the entire team.

Shopify's River data point — one in eight merged PRs co-authored by an AI agent — makes the organizational restructuring at Meituan and Ant Group look less like cost-cutting and more like a preemptive response to a world where shipping code is no longer the bottleneck.

The observation that PMs who modify code become Prototypers or Builders in practice, regardless of title, suggests that the real org chart in AI-native teams is already informal and archetype-based — the formal titles just haven't caught up yet.

Concepts & terms
Loop Engineering
An approach to AI-assisted development where autonomous loops — prompt, verify, iterate — replace step-by-step human prompting. The AI agent runs in a harness that cycles until a goal is met, rather than waiting for the next human instruction.
Shift Left
The practice of moving tasks earlier in a development pipeline. In Cherny's usage, it means pushing work traditionally done by specialized downstream roles (testing, operations) into earlier phases where agents or broader-role engineers handle it, compressing the waterfall.
Product-Market Fit (PMF)
The degree to which a product satisfies strong market demand. In Cherny's team-configuration model, a product's PMF maturity determines which archetypes the team needs most at that stage.
Iteration 0
The preparatory work before the first development sprint: setting up repositories, CI/CD pipelines, infrastructure, and architectural scaffolding. Cherny identifies this as one of the coding-adjacent tasks that current AI agents struggle to handle.
Source: juejin.cn ↗ Google Translate ↗ Backup ↗