Jun 26, 2025 · 1h 45m · mad
Guillermo Rauch: Why Software Development Will Never Be the Same
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Vercel CEO Guillermo Rauch joins FirstMark General Partner Matt Turck on The MAD Podcast to discuss the transformation of software engineering through vibe coding, autonomous agents, and the Model Context Protocol. He details Vercel's evolution into an AI Cloud infrastructure powered by v0, Fluid Compute, and isolated execution sandboxes.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 11.7% of the talking time here. How this is scored →
Matt as informed peer 4.0 Guest teaching 3.2 Guest disagreement 1.2 Matt pushing back 1.8
speaking balance: gold is Matt, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 1:37:41 Agent coding reframe
Guillermo pushes back on Matt's premise that aspiring engineers should still learn traditional syntax coding, arguing that agents will write code while humans must focus on system boundaries.
Hardest push from Matt ▶ 1:13:47 Unit economics challengeMatt directly confronts Guillermo with VC industry intelligence about AI coding platforms suffering from high user churn and negative gross margins, forcing Guillermo to defend v0's unit economics.
Biggest teaching moment ▶ 1:03:20 Serverless model shiftGuillermo educates Matt on why traditional cloud serverless billing fails for 15-minute o3-pro model reasoning requests, detailing how Fluid compute charges strictly for CPU execution cycles rather than idle duration.
Matt holds his own ▶ 1:13:47 Pressing on AI software marginsMatt demonstrates strong market expertise by citing competitor unit economics and pressing Guillermo on whether v0 is actually profitable on a per-generation basis.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Matt as informed peer | Guest teaching | Guest disagreement | Matt pushing back | Why |
|---|---|---|---|---|---|---|
| Vercel Ship Preview and Explosive Traction of v0 | 2 | 3 | 1 | 1 | Matt opens with general questions regarding v0 traction and team setup. Guillermo provides context and clarifies that the initial team started with 4 people rather than 10. | |
| Adopting the General Manager Model for Micro-Startups | 3 | 4 | 1 | 1 | Matt provides a clever framing of human organizational structure as microservices. Guillermo breaks down the market differences between vibe coding and agentic engineering. | |
| Defining Vibe Coding and Andrej Karpathy's Vision | 2 | 4 | 1 | 1 | Matt asks for a clear definition of vibe coding for broader listeners. Guillermo details Andrej Karpathy's vision and how v0 democratizes software development. | |
| Democratizing Coding and Infusing Models with Taste | 4 | 3 | 1 | 1 | Matt clarifies the definition of Next.js for the audience. Guillermo outlines how Vercel embeds developer taste and best practices into AI tokens. | |
| Full-Stack v0: Database Integrations and Enterprise AI | 5 | 3 | 1 | 3 | Matt challenges v0 as currently front-end centric and asks about backend ambitions. Guillermo acknowledges the framing and describes database and enterprise integrations. | |
| How to Become a Great Vibe Coder and Cultivate Taste | 2 | 3 | 1 | 1 | Matt asks how developers can cultivate taste and become effective vibe coders. Guillermo explains prompt enhancement and UI cloning from community templates. | |
| Replacing Traditional GUIs with Intent-Driven Agents | 5 | 3 | 1 | 2 | Matt introduces Benedict Evans's distinction between fixed GUIs and open AI interfaces. Guillermo expands on intention-driven agent architecture replacing frontloaded tool GUIs. | |
| Agentic Architecture and Automating Developer Pain | 4 | 3 | 1 | 2 | Matt asks Guillermo to clarify whether v0 is a copilot or a full agent. Guillermo explains how developer error loops forced an evolution into agentic architectures. | |
| Designing for AI Agents: Signal over Noise and CLI Resurgence | 4 | 4 | 1 | 1 | Matt asks how developer love is preserved when AI models make frequent errors. Guillermo explains how CLI tools and raw text formats like llms.txt serve AI agents better than visual UI. | |
| Technology Waves, Fighting AI Slop, and Latent Space | 3 | 4 | 1 | 1 | Matt asks how Vercel engineers practically cultivate taste. Guillermo analyzes historical tech wave anti-trends, contrasting AI slop with discovering latent space. | |
| AWS History and the Web 2.0 Mashup Origin | 4 | 4 | 1 | 1 | Guillermo recounts AWS history and introduces MCP as HTTP for AI agents. Matt grounds the abstract explanation by framing the Target.com consumer example. | |
| AI Cloud Security Primitives and Prompt Injection Threats | 6 | 3 | 1 | 2 | Matt accurately anticipates that open MCP architectures require specialized security and observability layers. Guillermo agrees and highlights prompt injection risks. | |
| Enterprise Adoption of AI Agents and Vercel Support Integration | 5 | 3 | 2 | 4 | Matt pushes back on enterprise willingness to adopt MCPs when it means giving up UI control. Guillermo counters with enterprise eagerness and internal support automation metrics. | |
| Unveiling the Vercel AI Cloud and Autonomous Operations | 6 | 3 | 2 | 5 | Matt directly confronts Guillermo on whether Vercel AI Cloud renders DevOps redundant. Guillermo clarifies the human-in-the-loop paradigm via pull requests and autonomous healing. | |
| Fluid Compute: Purpose-Built Serverless for AI Workloads | 5 | 3 | 1 | 1 | Guillermo details how 15-minute model reasoning loops broke traditional serverless billing. Matt summarizes the technical shift as a reinvention of serverless for AI workloads. | |
| Vercel Sandbox: Isolated Environments for Agent Code Execution | 3 | 3 | 1 | 1 | Matt asks about the Sandbox primitive. Guillermo explains sandboxing as the EC2 for model-generated code execution and details data flywheel feedback loops. | |
| Evolving Ideal Customer Profiles and Model-Forward Design | 5 | 3 | 1 | 1 | Matt asks how Vercel's Go-To-Market and ICP adapt as non-technical users adopt v0. Guillermo explains doubling down on dev-adjacent personas and model-forward UI design. | |
| Unit Economics, Margins, and Monetization in AI Software | 7 | 3 | 2 | 6 | Matt presses Guillermo on market rumors regarding poor retention and negative gross margins in AI coding products. Guillermo defends v0's positive margins and business model. | |
| Revenue Growth Drivers and the Virality of AI Tools | 6 | 3 | 1 | 1 | Matt cites specific financial figures showing Vercel's revenue growth from 100M to 180M. Guillermo attributes growth to AI SDK adoption and viral utility. | |
| Irreversible Efficiency Shifts and Mobile AI Control | 3 | 3 | 1 | 1 | Guillermo reflects on efficiency shifts using Einstein's observations on America. Matt adds lighthearted banter regarding European latency and efficiency. | |
| Public Communication, Product Simplicity, and Constructive Feedback | 3 | 2 | 2 | 1 | Matt asks Guillermo how he maintains public communication while handling negative feedback on social media. Guillermo emphasizes finding the kernel of truth in customer complaints. | |
| Work-Life Balance, Physical Fitness, and Startup Vitality | 3 | 2 | 1 | 1 | Matt asks personal questions regarding how Guillermo manages stress across family, angel investing, and running Vercel. Guillermo emphasizes physical health and startup vitality. | |
| Advice to Younger Self and Early Career Lessons | 2 | 3 | 1 | 1 | Matt asks what advice Guillermo would give to his younger self. Guillermo talks about building his first website and choosing non-obvious, low-reputation technologies early. | |
| Coding in the AI Era, System Limits, and Prompting | 5 | 4 | 2 | 4 | Matt challenges whether young people should still learn traditional coding. Guillermo reframes the question, asserting that agents will code while humans must understand systems and prompt effectively. |
Statements from this episode (46)
Rauch: Vercel's v0 has generated over 100 million applications
“We've passed over a hundred million generations of applications.”
Rauch: Vercel's v0 generates seven applications every second
“Every single second, there is seven app generations happening on V zero.”
Rauch: v0 doubled Vercel's entire user base in under a year
“VZero has more than doubled the entire user base of Vercel, and Vercel has been around for almost 10 years. VZero has doubled our number of users in less than a year.”
Rauch: Vercel built v0 with an original team of four people
“It was four people. And one was me who was doing QA when we were launching it.”
Rauch: Every company must become an AI-native interface or token factory
“Every company will have to rethink itself as what Jensen Juan calls a token factory, or what I could call, you know, a AI native interface, or you could call it an agent. It doesn't matter what you call it. The fundamental thing is that companies will be produ…”
Rauch: Next.js has 1.4 million monthly active developers as top React framework
“Next.js today is the largest web framework in the React ecosystem, 1.4 million monthly active developers.”
Rauch: Every company, task, and role will need complementary AI agents
“So as the market realizes that every company, every task, every role, We'll need these complementary agents.”
Rauch: Vercel's fluid compute platform reached 1 trillion function invocations
“We hit a trillion function invocations on our new fluid compute platform”
Rauch: Agentic engineering will be limited to professional developers until AGI
“Agentic engineering, I think for the foreseeable future, until we really, you know, reach SSI, or Safe Super Intelligence, or whatever you want to call it, or AGI, it'll still, it's going to be a more reduced number of software professionals.”
Rauch: The total addressable market for v0 is everyone with a computer
“When I think about the coding IDEs, You still have to be a coder, meaning you know how to download an ID, you have to, you know how to set up, you know, the dev server, the tooling, the SDKs, the libraries, whereas a product like vZero, the TAM is everyone wit…”
Vercel made its v0 code generation model available to external code editors
“One of the other announcements that we've made recently is we took the vZero model The sort of underlying code generation engine, and we've made it available to code editors as well.”
Rauch: Midjourney, Walmart, and Nike are built with Next.js
“Midjourney.com is built with Next.js, but so is Walmart.com and so is Nike.com.”
Rauch: Taste, vision, and clear articulation are essential for AI creators
“You need taste. You need to have a vision of what you want, and you need to be able to articulate it. You know, those are like the fundamental, like literacy of a person. A successful individual will be able to discern what they want and what they don't want.”
Rauch: AI agents, not human users, will directly wield software tools
“So we're entering a world where it used to be that the tools get drawn in front of your face and knew how to wield them. In this new world, the tools are given to the agent and the agent wills them and decides when to use which.”
Rauch: Ambitious AI coding requires agentic architectures due to LLM errors
“As you get more ambitious with the things that you wanted to do, like create more interactive code, data fetching code, et cetera, you always end up with an agentic architecture because you realize that LLMs are not bulletproof.”
Rauch: v0 suppresses error messages when AI can automatically fix them
“Sometimes in V zero, we don't even render the error because we know with a high probability that the agent is going to fix it.”
Rauch: Dev tools must be designed for both humans and AI agents
“LLMs.txt or raw markdown is better for agents. So if you're building developers still today, you have to have that duality in your head. You need to think, okay, How do I make my content, my errors, my developer tools great for agents?”
Rauch: Anthropic's Claude Code uses Vercel to deploy applications
“But nowadays when you ask Claude Code Anthropics agent to deploy, They use for sale because we had, you could argue accidentally created the perfect tool for an agent to deploy.”
Rauch: Developing product taste requires maximizing 'exposure hours' to products
“And so what I call exposure hours is you have to, one, you have to understand how people interact with your own products today. If you already have a product, You have to really pay attention to, okay, how are people exactly leveraging your product? Because I …”
Rauch: AWS started as e-commerce APIs, not spare hardware capacity
“A lot of people believe that AWS started out because it had spare hardware. If you listen to the AWS episode on the acquired podcast, the story is actually way more interesting. The reason that it's called Amazon web services. Is that there was a hypothesis in…”
Rauch predicts a fully generative internet with fluid interfaces within 10 years
“So I think what's going to happen over the next 10 years is that people are going to start creating cross pollinations between products and integrations between products and novel interfaces to products that weren't, I mean, they were possible before, but just…”
Zapier and Solana host their Model Context Protocol servers on Vercel
“We have companies like Zapier using Vercel to host their MCP servers, and companies that are, like, doing significant volume, by the way, which actually kind of surprised me. Solana, for example.”
Rauch: MCP will replace human business development with agent-to-agent meetings
“You know, in many ways, I think MCB will be the new business development, but it's going to be at a hundred times the speed. It's not going to be people meeting and it's going to be agents meeting.”
Rauch: Multiple security incidents have already occurred with MCP servers
“In fact, there's already been quite a few incidents with MCP servers where depending on what chat message was sent, the MCP tool that was called was made to disbehave.”
Rauch: AI agents and MCP require net-new security infrastructure and tools
“So it's a brave new world and new security infrastructure and tools are required.”
Rauch: Web experiences will evolve to be hyper-personalized starting with search
“I think the web will evolve to have this hyper personalized experience. It's probably going to start with search.”
Rauch: Agentic commerce will feature AI agents navigating alongside web users
“And so I think there's going to be new primitives like that in the world of agentic commerce, where the agent jumps alongside you in your navigation journey.”
Rauch: Vercel's support team uses v0 to suggest product improvements
“Our customer support team uses VZero very extensively because they want to produce suggestions of, hey, if we make this part of the product better, if we improve this area of this, of the service, we will avoid this category of issue that is arriving to me.”
Guillermo Rauch: Vercel hosts OpenAI.com, a top 20 web traffic site
“Like we host openai.com is one of the top 20 websites on the internet.”
Rauch: AI cloud will be self-healing and autonomous with humans in loop
“We believe that the AI cloud will be self healing as well as completely autonomous from an infrastructure point of view. And we believe that agents and humans will collaborate.”
Rauch: Vercel suffered no downtime during major Google Cloud outage
“You saw the Google cloud outage the other day for hours, a lot of the cloud was down. So I'll give you an example. Vercel didn't go down”
Rauch: Vercel Fluid Compute only bills active CPU cycles during LLM wait times
“If you're waiting for 15 minutes for OpenAI to respond, and you do a tiny bit of compute, maybe to like transform it into HTML or transform it into an MCP response or, you know, call other systems. We're only going to charge you for the actual CPU cycles that …”
Rauch: Vercel Sandbox is the Amazon EC2 of AI
“So you can think of sandbox as the Amazon easy two of AI. It's not for the code that your developers wrote is for the code that gets emitted by the LMS and it allows you to create this incredible new products.”
Rauch: AI product builders must design from current model capabilities forward
“And there's another emergent principle of you have to design from the model forward.”
Rauch: Vercel v0 traction improved 100x following interface redesign
“We call it internally V one, which is kind of a joke, but when V one came out, our traction improved by a hundred X.”
Rauch: Vercel's v0 has positive and improving gross margins
“Yeah, BZero has positive gross margins. It's a healthy business. The margins are improving, the business is improving”
Rauch: Adding monetization features to v0 will improve user retention
“And so I think V zero, you'll see same kinds of things come to V zero so that I think what's going to improve retention is that you're not just like vibe coding into the void. You're actually creating a successful economy.”
Rauch: Vercel AI SDK is second largest AI module behind OpenAI
“It's the second largest AI module behind OpenAI.”
Rauch predicts a huge wave of mobile apps controlling remote computers via AI
“Because now you can control computers remotely with agents. And again, we're just at the beginning of that. I think there's very few people that have actually done this. There's a company called Manos AI that has like this, like on your phone, you are controll…”
Rauch: ChatGPT's UI simplicity gives it a fundamental advantage over Google
“If you look at ChatGPT, one fundamental advantage they have over Google is the simplicity of the UI.”
Rauch: There is a positive correlation between prompting and success
“I do believe that in life there is a positive correlation between prompting and success.”
Guillermo Rauch: Large tech companies always compete with garage vibe coders
“It doesn't matter how big your company is, how much history you have, how much money in the bank you have, you're still competing with the seed of an idea of a vibe coder in a garage somewhere.”
Rauch: Career alpha comes from choosing emergent tech over high-status trends
“Where I found the alpha in my life was in actually not immediately go to the thing that was like the obvious high reputation thing. When I started using JavaScript, it was considered to be toy programming language. And it couldn't be fast and it couldn't be co…”
Rauch: Prompting AI is the best proxy for measuring candidate curiosity
“I think prompting is the biggest proxy for how would I measure someone's curiosity.”