Guillermo Rauch on v0 and the Future of Building

Guest: Guillermo Rauch — founder and CEO of Vercel; creator of Next.js and Socket.IO.
Host: Lenny Rachitsky
Source: Lenny’s Podcast.

Overview

Guillermo Rauch built Vercel to host the web’s most demanding applications, and v0 to expand who can build them — from 5M React developers to 100M “product builders.” The conversation covers the democratisation of product building, which skills survive the AI transition, how to develop taste, and Vercel’s internal operating principle of increasing exposure hours.

Key ideas

Everybody can cook. v0’s mission, borrowed from Ratatouille: remove the barrier between having an idea and shipping it. The addressable market for product building is not developers — it is every person who talks to others about what they want to see in the world.
Translation tasks are going away. The historical specialisation of converting a design into CSS/Tailwind/React implementation is a translation task — the same type of task that LLMs were originally built to solve (Google Translate). These specialisations are being automated.
Knowing how things work vs. knowing everything. Deep encyclopaedic CSS knowledge goes away; conceptual understanding of systems and eloquence (knowing the tokens: “turbulence”, “sepia filter”, “neobrutalist”) becomes more valuable for steering models toward intent.
Exposure hours as taste. Vercel’s internal operating principle: quantify how much time you spend watching real users interact with products. Taste is not innate — it is built through exposure, feedback, and iteration.
Code last, not code first. The future of product building: start with experience (what you want the user to feel), let AI handle implementation, keep code as an escape hatch for precision and getting unstuck.

The “everybody can cook” thesis

Guillermo’s back-of-the-napkin TAM calculation:

React developers: ~5M
JavaScript developers: ~20M
People who talk to others (e.g., on Slack) about products they want to build: ~100M (Slack’s MAU as a proxy)

v0’s mission: close the gap between having a product idea and shipping a real product, for the 100M, not just the 5M. “Social product building” — the next evolution of GitHub, but with no prerequisite coding skill.

Translation tasks

“A lot of the programming jobs to be done that used to be specialisations, I think, are going away. They’re translation tasks.”

The connection: transformer architecture originates from neural machine translation (Google Translate). CSS/Tailwind styling from a design is structurally the same task — converting intent in one representation (design/screenshot) to implementation in another (code). LLMs do this natively.

Guillermo’s test: every model generation, he tries to re-create his own website with v0. With the last model, it took 10–15 prompts; with the current model, 2 prompts. As a 25-year expert front-end engineer, he finds v0 generates more accessible code than he wrote.

Skills for the future

Skills that persist (Guillermo teaches these to his five children):

Mathematical/logical reasoning — fundamental understanding of how things work.
Eloquence — knowing the right tokens to steer models: not just “make it nice” but “neobrutalist”, “sepia filter”, “turbulence effect”, “newspaper-like”.
Conceptual understanding of systems — not the encyclopaedic detail but the meaningful map of what exists and how it interacts.
Presenting and putting yourself out there — in a world where marginal cost of software drops to near zero, audience and brand become critical.

Skills that go away:

Specialised CSS property knowledge (knowing when each browser supported each property).
Layout/styling translation from design.
Most surface-level programming specialisations.

Foundational infrastructure engineering survives: LLMs orchestrate tools but cannot write the compiler from scratch. The engineering that LLMs depend on remains human-built.

Developing taste: exposure hours

Vercel’s internal operating principle:

“Try to quantify how much time you expose yourself to watching how people use your products.”

Taste is not innate. It is a skill built through:

Exposure to many products. Use the best products; try the worst; notice what distinguishes them.
Shipping and getting feedback. Build things, give them to users, watch where they break and succeed.
Customer meetings. Guillermo reserves a third of his meetings for customers, watching them use the product live.
Demo Fridays. Vercel has a ritual every Friday where people demo what they built with AI tools — creating space for unexpected generalism.

The inertia is always toward thinking you know everything. Exposure hours is the antidote.

Code last, not code first

“You’re living in the product.”

The v0 design philosophy: experience first, code last. Start with the front end (what the user sees and feels), not the back end (how it works). Code is an escape hatch — always visible, always editable, always available to hand off to another AI or another engineer when stuck.

Escape hatches as design principle (borrowed from React’s API design): when a tool cannot perfectly model your problem, give users a principled way out. v0 shows code; users can edit it directly, copy-paste to other models (e.g., Claude, o1), or switch to traditional git-based development.

Getting unstuck: just say “try something else.” The model treats it as legitimate feedback and retries with a different approach.

v0 in practice

1.3M+ users; v0 Community has 20,000+ submissions in under a month.
Community model: fork others’ starting points instead of starting from scratch — same spirit as open source.
Use cases Guillermo has shipped: flight radar (built during a Japan Airlines flight with poor WiFi, <2 hours, $20/month subscription cost).
Enterprise customers: some organisations have moved entirely to v0 for all product and client work.
Key limitation: large codebases are harder for AI to reason over; scope tasks to specific files or components.

AI at Vercel

Design team generates animations with video models.
Marketing team creates infrastructure demos with v0 (replaced static diagrams).
VP of Sales Engineering built prospect education tools with v0.
Product managers build live interactive PRDs instead of static documents — the entire team says “just ship it.”
Vercel uses v0 to build v0 and Vercel’s own infrastructure.

Vision

“I just see a future where AI becomes synonymous with software. We build software and we use software to build software.”

The endpoint: stop talking about AI as a category. AI is software. The company is a software company that builds software using software.