Product Builder vs PM: What the Role Is and How to Get There

🌏 中文版

Anthropic's Boris Cherny, the creator of Claude Code, said something worth sitting with:

Today coding is practically solved... We're going to start to see the title of 'software engineer' go away. It's just going to be 'builder' or 'product manager.'

This isn't a prediction. It's already happening.

In June 2026, Boris unpacked that statement into a concrete framework. Observing how the Claude Code team actually works, he found that roles aren't divided by job title — they operate along five archetypes:

1. Prototyper: constantly generates brand-new ideas; most won't ship
2. Builder: quickly turns a surviving prototype into production-grade product or infrastructure
3. Sweeper: cleans up the UI, simplifies the code and system, unships unnecessary features, optimizes performance
4. Grower: takes a shipped product and iterates on it to improve Product-Market Fit
5. Maintainer: owns a mature system's security, reliability, speed, and efficiency

Boris emphasized that many people span 2–3 archetypes, and these archetypes are not tied to job function — at Anthropic, some designers are Prototypers, some are Sweepers; the same spread appears among engineers, PMs, and data scientists. He also mapped archetype mixes to product stages: pre-PMF needs heavy 1+2+3, growth needs 2+3+4 plus some 5, and mature products run on 3+4+5 with some 2.

Aakash Gupta's follow-up analysis added the cultural context that makes this work at Anthropic: everyone shares the title Member of Technical Staff, they don't write PRDs — they build first, most prototypes are expected to die (the Claude Code spinner animation went through 50–100 iterations with ~80% never shipping), Claude Code reviews every PR before a human does the final check, and there's a framed copy of The Bitter Lesson on the wall where the team sits.

The tweet accumulated over 3 million views and sparked extensive discussion across Chinese and English-language communities (see references at the end). This article approaches from the broader "Product Builder" role and places Boris's five archetypes within the industry context.

What Is a Product Builder

A Product Builder is someone who can take an idea from concept to a working product with minimal dependency on other teams. Not a PM, not a designer, not an engineer — but capable of doing all three.

Traditional product development is an assembly line: PM writes the spec → Designer produces mockups → Dev implements → QA signs off. Each handoff carries waiting time and communication overhead. Product Builders compress this pipeline into a loop — one person who can rapidly validate assumptions and iterate on solutions.

The core distinction: PMs influence team execution through authority; Product Builders directly contribute output through capability. A PM's deliverable is a PRD or roadmap. A Product Builder's deliverable is a working prototype or feature.

Why Now

Two words: AI.

In February 2025, Andrej Karpathy introduced the concept of vibe coding — instead of writing code line by line, you describe what you want in natural language and let AI generate it. This directly lowered the barrier to building things. For the concrete patterns this way of working has settled into, see the 190 patterns in the Encyclopedia of Agentic Coding Patterns.

A McKinsey experiment with 40 product managers found that generative AI raised individual PM productivity by 40% — while product time-to-market improved by only 5% in the same study.

That gap is worth sitting with. Faster individual output does not mean a faster product line. What usually blocks a release is not "we can't build it" — it is decision-making, alignment, and validation. That is precisely the stretch the Product Builder role claims to collapse. It is also a warning: drop AI into an unchanged process and the gains get eaten by the handoffs.

Adoption is slower than the headlines suggest, too. The widely quoted "78%" from GitLab's 2024 Global DevSecOps Report covers respondents already using AI or planning to within two years; in the same report, only 26% said they had actually implemented it.

When tools like Claude Code, Cursor, Lovable, and Replit let one person go from idea to working prototype in a few hours, the traditional trio of PM + Designer + Dev is no longer the only option.

LogRocket ran the numbers: a traditional three-person product team costs roughly $1.2–1.5M per year, and 50–60% of shipped features underperform expectations. If a single Product Builder can validate assumptions before committing full engineering resources, avoiding just five unnecessary features per year saves over $500K.

The math is persuasive, but it only counts the savings. At least three costs never make it into the denominator:

One: the security and maintenance cost of AI output. The calculation assumes what a Product Builder ships counts cleanly as output. But as covered below, Veracode measured OWASP Top 10 flaws in 45% of AI-generated code. Some of those saved engineering weeks come back as security review and technical debt.

Two: the expected cost of worse validation. The math assumes one person validates as accurately as three. Strip out the design and engineering perspectives and the assumptions themselves get narrower — you validate the wrong problem faster, and the savings turn into the cost of going the wrong direction.

Three: outcomes may not improve at all. SVPG's observation is the most direct version of this: teams deliver faster with AI and their outcomes do not improve. If what accelerates is output rather than judgement, the $500K is saving the cost of building the wrong thing — or forfeiting the chance to build the right one. The math does not answer which.

The honest financial argument for this model is not "one person is cheaper than three." It is "under what conditions is one person's judgement not worse than three people's." That is an operating question, not a headcount question.

The Market Right Now: Who's Hiring, How They Define It, What They Pay

Big Companies Leading

This isn't experimentation at small companies — large organizations are already moving:

• Amazon (Ring & Blink) stripped all traditional job titles during annual reviews. Hundreds of employees became Builder and Builder Lead. CPO Jason Mitura's internal memo (obtained by Reuters): "We define and reward success through one question: what is the scope and magnitude of the customer value you create?"
• Stripe made PMs builders earlier than most — new PMs get the latest models and internal coding agents on day one. But they're already asking the next question: when engineers can spin up fifty agents overnight, what's the point of a PM proudly shipping PRs? Kevin Yien (Stripe PM lead): "It's not one function eating another. It's every function collectively moving over so that we can all actually get more done"
• LinkedIn ended its APM program and stood up an Associate Product Builder (APB) track. CPO Tomer Cohen at Miro Canvas 26: "We're basically imagining this future for our customers, but internally we're still slotted into PM, design, engineer — the lines do not make sense anymore"
• Walmart opened Agent Developer roles (self-described as its first "biz/tech" position), no technical background required. Chief people officer: a year ago we had no dedicated agent builders, today we do
• Meta went beyond self-description — a 1,000-person team within Reality Labs formally restructured, leaving only three titles: AI Builder, AI Pod Lead, AI Org Lead. PM Jeremie Guedj (10+ years as PM) went public first in February 2026
• Google — the head of product for AI Studio and Gemini API, Logan Kilpatrick, dropped the PM title for Member of Technical Staff. But context matters: DeepMind is simultaneously hiring PMs with traditional titles at $183k–$320k base. One person changing a title is not the same as a company restructuring
• Cursor — 40 engineers, 1 PM, 0 dedicated designers, past $4B ARR. OpenAI Codex is similar: 43 people covering 10–12 product surfaces that would need 150–240 at a traditional company
• Cars24 (Indian used-car platform) eliminated all job titles, grades, and org bands company-wide. Everyone is now Builder. Co-founder Vikram Chopra: "Every generation gets to question one assumption that previous generations took for granted. Ours may be the assumption that companies need to be organised the way they have been for the last hundred years." Revenue per employee up 50% and nearly 300bps of EBITDA improvement over 18 months
• Block (Square's parent) started calling some managers player-coach — not the Builder label, but the same direction: managers who also build
• PayFit defined the Product Builder role back in 2019, using their in-house low-code language JetLang — nothing to do with the AI wave

One Title, Different Jobs

Line these up and something rarely pointed out becomes obvious: "Product Builder" does not describe the same role across companies.

LinkedIn hires people with no product experience yet. Mews hires senior engineers. Walmart explicitly requires no technical background. PayFit's has nothing to do with AI. Asking which department a role grew out of tells you more than the title does.

July 2026 Job Snapshot

Positions actively hiring at the time of writing. Postings close — this captures the market as-is, not a permanent directory.

International

Taiwan

What the Listings Show

ShipBob is the most instructive case. They built an entire job family called "AI Builders" — its own career ladder (Director of AI Product Builder), pod structure (AI Engineer + AI/Prompt Engineer + Designer + AI Product Builder), and operating model (discovery → spec → prototype → production PR, one person end-to-end). This is the most institutionalized version on the market.

Unicorns are adopting the title. Abnormal Security (cybersecurity) and ShopMy (creator commerce, $1.5B valuation) are both hiring Product Builders at $141k–$225k. This is no longer a small-startup experiment.

Taiwan already has 5 open positions, with the same spread as the international market: CMoney's is new-grad training (mirroring LinkedIn's APB), 菜蟲農食's is forward-deployed hands-on, 光時代's is essentially an AI-native engineer.

India is especially active. Optima, HighLevel, and others are hiring in India with compensation from ₹30k/month to ₹1Cr/year. HighLevel's "Team of One" is the most extreme framing.

The salary range reflects the definitional divergence: ShipBob senior goes up to $290k; Foundever junior requires only 1-3 years. ZipRecruiter's US average of $159k is directional at best — the title covers different jobs.

Khan Academy's Sal Khan put it plainly:

The people who are just waiting to get the spec... they're going to have trouble. But the people who are like, 'I'm going to go meet with the customer, and I can build it,' I think they're going to do great.

What Skills Does It Take

Product Builders don't need to master everything — just know each domain well enough to move independently:

Technical: Basic Python / JavaScript, API integration, ability to write serviceable code with AI assistance. Not a senior engineer, but capable of AI pair programming.

Design: Using v0, Lovable, Claude Design, or Claude Artifacts to produce interactive prototypes, and judging UX at the working-code level — not in static mockups.

Product: User research, data analysis (SQL), hypothesis validation, prioritization.

AI literacy: Prompt engineering, understanding the capabilities and limitations of AI tools.

That last one is not a bonus, it is the threshold. Veracode's 2025 GenAI Code Security Report tested over 100 models and found that 45% of generated code introduced an OWASP Top 10 security flaw. Java failed 72% of the time; XSS went undefended in 86% of the relevant samples. The most telling finding: models kept getting better at producing code that runs, and barely improved at producing code that is secure.

In other words, a Product Builder's technical judgment is not there to write faster — it is there to notice what the AI got wrong. This is the most underestimated cost of the role.

What the Skeptics Say

Everything above is the upside case. The role also carries real costs, and the people raising objections are not outsiders.

Roman Pichler's rebuttal deserves the most serious attention. His point is not "one person will burn out." It is that when one person can do all the work, distributed intelligence disappears from the process. The fewer disciplines involved early, the more you are simply moving faster on narrower assumptions. Fast validation only pays off if the assumptions are good — and assumption quality usually comes from friction between different perspectives. That friction is exactly what gets compressed away.

SVPG's observation is harder to argue with. In "AI Product Management 2 Years In," Marty Cagan notes that the teams they see are genuinely delivering faster with AI — and their outcomes are not improving. That points at the same thing the McKinsey numbers did: 40% individual productivity, 5% time-to-market. What accelerates is output, not judgement.

Practitioners have their own doubts. In Userpilot's 2026 survey:

• 46.7% worry about being asked to do too much with too little support
• 37.3% fear ending up in a state of perpetual role confusion
• 31.6% worry that doing everything means doing nothing particularly well

There is also a practical side effect: without clear boundaries, a Product Builder easily steps into design and engineering territory, undermining teammates' autonomy until nobody is sure who owns what. The confusion bought with that speed sometimes costs more than the time it saved.

The More Likely Outcome: The Role Splits in Two

A better description of what is actually happening than "every PM must become a builder" is this: the role is forking.

• Builder PM: AI-native, prototypes independently, low dependency on other teams. Fits early exploration, internal tools, fast iteration
• Integrator PM: strong at alignment and communication, keeps marketing, sales, and product pointed the same direction. Fits complex organizations and scaling stages

Both are growing. Neither is replacing the other. What is actually getting squeezed is the middle — the role that only moves information, writes specs, and maintains roadmaps, neither building nor owning alignment. That is the real situation of the "non-creator PM" Cagan warns about.

So the question is not "should I become a Product Builder," but "which side am I moving toward, and can I get deep enough on it."

When This Model Actually Fits

When product complexity increases, when you need large-scale system architecture, or when deep user research is required, specialized roles remain irreplaceable. Product Builders are best suited for:

• Early-stage products that need rapid exploration and validation
• Internal tools that don't require large-scale engineering investment
• Feature iteration that calls for rapid experimentation and data-driven decisions
• Any "validate before committing" phase

LinkedIn's Aneesh Raman said it well:

The full stack builder takes what would've been days or weeks as a conveyor belt between design, product, engineering... and gives it to an individual with these tools.

What the Chinese-Language Discussion Looks Like

For readers who also follow the Traditional Chinese product community, this topic is already well covered there — and its slant is worth noting. Peter Su treats the role as the natural result of AI collapsing the cost of turning an idea into something concrete; AAPD in Taiwan already runs a course on building Product Builder skills, approached from a designer's perspective; CMoney writes from hands-on AI Lab experience, arguing that what separates people is not skills but problem-framing and user understanding; and Vista focuses on PM anxiety itself.

One observation: these pieces lean consistently optimistic, treating the role as opportunity rather than trade-off. The counter-arguments in the section above — Pichler on distributed intelligence, SVPG on outcomes — are almost absent from that conversation.

If You Want to Move in This Direction

Whether you're currently a PM, designer, or engineer, the path is the same: fill in the gaps you're missing. Worth noting: LinkedIn's APB program explicitly welcomes a career pivot — no formal product experience required, open to people already working full time. That alone says the entry point to this career path is wider than traditional PM.

PM → Pick one real problem from your backlog. Build 5 different approaches in Claude Code or Codex, 30 minutes each. Kill 4, keep 1. This is Prototyper muscle memory. You're not learning to code — you're learning to replace meeting-room arguments with something that runs.

Designer → Use v0, Lovable, Claude Design, or Claude Artifacts to turn your designs directly into interactive prototypes. Judge UX at the working-code level — does the layout hold up, does the flow feel right, is the feedback loop fast enough? Anthropic's designers already submit PRs. Static Figma handoffs are no longer the deliverable.

Engineer → Do 3 fifteen-minute user interviews this week. Watch 5 session recordings. Then claim ownership of one user-facing metric. Not "spend time understanding users" — turn customer signal into a number you check daily.

Product Builder isn't a job title — it's a way of working. In an era where AI enables everyone to do more, people who can independently move from problem to solution will become increasingly valuable.

References

• 'Engineer' is so 2025. In AI land, everyone's a 'builder' now - SF Standard
• Why product managers must become product builders in 2026 - LogRocket
• AI is turning product managers into builders - Fast Company
• What It's Like to Be a Product Builder in 2025 - CuriousCore
• The Vibe Coding Imperative for Product Managers - ACM
• Introducing the Product Builder - PayFit
• The Era of the Product Creator - SVPG — Note: Cagan and Baxley argue here that the PM role remains necessary and should work shoulder-to-shoulder with designers and engineers daily; their warning is aimed at non-creator PMs. That sits in tension with this post's framing of one person compressing the whole line, and is listed for contrast
• What It Means to Be a Product Builder at Mews - Mews Developers — one of the few companies placing the role explicitly on an engineering track
• Comment le métier de Product builder est apparu chez PayFit - PayFit — PayFit's Product Builder role and the JetLang low-code platform
• Should Product Managers be Product Builders? - Roman Pichler — the most substantive skeptical case; its core argument is the loss of distributed intelligence
• AI Product Management 2 Years In - SVPG — teams deliver faster with AI while outcomes fail to improve
• 6 Product Management Trends in 2026: The PM Role Is Splitting - Userpilot — the Builder PM / Integrator PM fork, plus practitioner concern data
• How generative AI could accelerate software product time to market - McKinsey
• 2024 Global DevSecOps Report - GitLab
• 2025 GenAI Code Security Report - Veracode
• Why LinkedIn is replacing PMs with AI-powered "full-stack builders" - Lenny's Newsletter
• All in on Agents - Walmart Global Tech
• How Walmart is securing a new cohort of AI builders - IT Brew