ACQ2: Building Web Apps with Just English and AI (with Vercel CEO Guillermo Rauch)
Key insights
Media referenced
- Open Operator - other - Open-source Vercel-customer-built clone of OpenAI's Operator agent, launched within 24 hours of Operator's release using v0, cloud browser infrastructure, and DeepSeek models.
- OpenEvidence - other - A 'v0 for medicine' product Rauch cites as an example of companies building domain-specific AI copilots with Next.js and Vercel.
Companies
- Vercel - Rauch's company; infrastructure platform for building and deploying web applications, maker of Next.js, v0, and the AI SDK; ~$100M ARR as of May 2024, growing 80% YoY, 550 employees.
- Next.js - Open-source React framework Rauch created; now has 1.3 million monthly active developers and underpins most major AI products (ChatGPT, Claude, Sora).
- Meta - Open-sourced React, the low-level UI library Next.js builds on top of; internally had its own Next.js-like layer that was never released.
- Google - Open-sourced Kubernetes, cited as the complex infrastructure tooling Vercel was built to abstract away; Google Translate cited as an analogy for AI displacing a human specialization.
- Socket.IO - Real-time communication library Rauch built prior to founding Vercel, reflecting his early obsession with dynamic, real-time web experiences.
- Polymarket - High-traffic prediction market built on Vercel, cited as proof a tiny team can sustain massive infrastructure load without a dedicated infra team.
- OpenAI - Built ChatGPT and Operator on Next.js; a Vercel customer cloned Operator's functionality as 'Open Operator' within 24 hours using v0.
- Databricks - Cited alongside Snowflake and Datadog as a peer business model: open-source core adoption plus a highly optimized, usage-priced commercial compute layer.
- Snowflake - Cited as a peer usage-based business model where customers pay for what they use rather than a flat license.
- GitHub - Cited as the closest prior analogy to Vercel's model - builds on open-source Git and monetizes URLs to code, as Vercel monetizes URLs to working applications.
- Red Hat - Cited by David Rosenthal as the earlier, weaker open-source business model (pure support/services) that Vercel's model improves on.
- Solana - Blockchain whose web properties are hosted on Vercel; crypto traffic and attacks are credited with hardening Vercel's DDoS and security infrastructure.
Techniques and frameworks
- Vertical integration of framework and infrastructure - Rauch's core strategic bet: pairing an open-source front-end framework (Next.js) with a proprietary, highly optimized hosting layer (Vercel) rather than picking one side.
- Infrastructure as an output of the application - Vercel's inversion of the typical cloud approach - developers declare the application in Next.js and infrastructure (serverless functions, CDN, regions) is generated from that, rather than provisioned first.
- Preview deployments / ephemeral environment URLs - Every git branch gets an automatically deployed, shareable URL, replacing shared staging machines and in-person demos with async, cross-functional collaboration.
- Domination of a single niche before expanding - Rauch's company-building advice: be world-class at one narrow thing (front-end) before letting customer demand pull you into adjacent categories (databases, back-end), rather than launching broad.
- Guardrails for AI code generation (evals, benchmarks, RAG, custom design systems) - How Vercel encodes its engineers' taste and best practices into v0's outputs so AI-generated code and design meet a quality bar instead of producing 'slop.'
Summary
Guillermo Rauch, founder and CEO of Vercel, walks Ben Gilbert and David Rosenthal through the company's decade-long arc from a scrappy infrastructure bet to a ~$100M ARR platform now being reshaped by AI. He frames Vercel's founding insight as contrarian: while the front-end ecosystem chased static site generation in the mid-2010s, Rauch bet on cloud-rendered, dynamic applications because it offloaded compute burden from an increasingly mobile-dominated device landscape. That bet produced Next.js, the open-source React framework he wrote to give developers "the car" to match the open-sourced "engine" of React and Kubernetes, and Vercel, the proprietary infrastructure layer built to run it at scale.
A recurring theme is Vercel's specific open-source strategy: give away the programming model (Next.js, the AI SDK) because that's where open source adds real leverage, but keep the operationally complex hosting layer closed, because "open source hits diminishing returns" once you're talking about global regions, DDoS mitigation, and rolling upgrades. Rauch draws direct parallels to Databricks, Snowflake, and GitHub as businesses that monetize a highly optimized proprietary layer sitting on top of an open or widely-adopted component, contrasting this with the weaker, support-only open-source business models of the Red Hat era.
The conversation's center of gravity is v0, Vercel's AI web-development assistant, which Rauch describes as growing faster than anything he's experienced: $1M ARR in 10 months, then doubling to $2M in 14 more days, then to $4M twenty days after that. He's candid that v0 now produces front-end code and design he judges better than his own hand-written work, including on accessibility, and frames this as the leading edge of a broader shift where AI takes over "translation of intent into working code" the way machine translation displaced professional translators. Vercel's response has been to encode its own engineers' taste into v0 through evals, benchmarks, RAG, and custom AI-tailored design systems, so the AI's output meets a deliberate quality bar rather than producing "slop."
Rauch also discusses company-building philosophy: his advice to founders is to dominate one narrow niche before letting customer demand pull you into adjacent product categories, rather than launching with a product that matches the full size of your ambition. He credits Vercel's discipline here (staying tightly coupled to Next.js and front-end for 6-7 years before expanding into back-end and databases) as core to reaching its current scale with only 550 employees. He describes an internal practice of denying the v0 team any special infrastructure accommodations, as a constant proof point that the platform has "no graduation risk" for customers either.
Crypto comes up as an unexpected forcing function: the sector's high-traffic, high-attack-surface workloads pushed Vercel to build industry-leading DDoS response times and secrets/wallet encryption, hardening that now serves the Fortune 500 banks and large enterprises increasingly adopting Vercel. Rauch closes on the implications of Next.js standardization for M&A (acquired companies integrate faster because their front-ends share a common shape) and on his own philosophy of CEO involvement: staying hands-on across every function by using Vercel and v0 himself daily rather than relying on executive briefings.
Notable Quotes
"It can create code better than what I can do by hand." - Guillermo Rauch
"We developed the last framework so companies would not have to do their own adventure." - Guillermo Rauch
"Vercel gives you URLs to working applications." - Guillermo Rauch
"As a CEO you have to understand. You can't just say, well this area of infrastructure, I have no idea how it works. Talk to Johnny." - Guillermo Rauch
"Marginal cost of software production goes to zero." - Guillermo Rauch