Ad Narrator | Sourcery Summary

1. Key Themes

Agents Have Become the Primary Deployer of Software

Vercel's usage data shows the shift from human-written to agent-written software happening at a speed the builders of AI themselves didn't anticipate. Rauch frames Vercel as the "universal field" where coding agents push the code they write. Deployments that took a decade to reach 1 billion added another 1.5 billion in roughly 10 months.

Rauch: "We started out the year with 3% of deployments on Vercel were coming from coding agents. Now, 60% are agentic." 00:00:18

Rauch: "It took us about 10 years to get to a billion deployments... And then in the next 10 months, the 10 months since then, until now, we've added another 1.5 billion deployments." 00:00:32

Rauch: "Vercel has sort of emerged as the place, this universal field where all of the coding agents are plugging into and pushing their code out so that it can be used by anybody." 00:10:11

The Web Is Being Rebuilt for Agents, Not Humans

Rauch's central thesis is that the internet is moving from machine-to-human to agent-to-agent. Companies must now present "agent-readable" versions of their business: straight signal rather than pixels, MCP tool exposure, and new commerce protocols. Vercel is doing this for itself and helping customers do the same.

Rauch: "The era that we're going into now is more agent to agent... When an agent goes to vercel.com, I don't give them the fancy... beautiful triangle with a gradient. The agent don't care. The agent wants straight signal." 00:18:23

Rauch: "The internet for humans was HTML. The internet of agents is new protocols. One is called MCP, which basically allows you to say, 'Hey, agents, these are all the tools that you can use from my service.'" 00:25:14

Rauch: "There's new protocols also emerging. For example, one is the Machine Payment Protocol or the Universal Commerce Protocol." 00:25:42

Rauch: "Funny enough, a lot of the internet was built to block bots... Vercel is actually taking the opposite posture. We're saying, 'No, no, no. The moment is now to help every business in the world retool themselves for agents.'" 00:20:29

Agents as Purchasers and "Bot Intelligence" Replacing Bot Detection

Vercel is seeing agents transact autonomously, buying services from its marketplace with no human intervention. This turns agents into a new customer segment that businesses must serve and measure.

Rauch: "We see agents now purchasing our services... if an agent needs to buy a web browser, they can purchase BrowserBase through the Vercel marketplace. And I'm enabling that transaction end-to-end without any human intervention." 00:22:06

Rauch: "What you do need that we provide is bot intelligence. It's so important to know if the bot is having good experience, which is kind of nuts to think about, right?" 00:21:42

Personalized, Generated Software vs. Aggregated SaaS ("Yes-Code")

Rauch takes an opposing stance to the roll-up model of buying and aggregating software. If code is nearly free, enterprises should generate bespoke, domain-specific tools. The lasting value of incumbents like Salesforce and SAP shifts to being headless systems of record.

Rauch: "I don't believe in sort of aggregating piles and piles of software... I believe in a future of personalized software." 00:30:08

Rauch: "I'm a yes-code company, because I'm now saying coding agents basically make the code free." 00:30:37

Rauch: "The lethal combination there is that you buy Vercel as the platform, and then you never have to buy software again, because you can just generate it to your needs." 00:00:32

Rauch: "The sources of, uh, record... that's critical data infrastructure. I'm not going to call it software... headless means that we no longer go to the head ourselves. The agent can go and get the data out of the system." 00:31:41

Generative Interfaces Replace Product-Designed Screens

Internal tooling at Vercel is already agent-first. Instead of a product team deciding what fixed UI an employee sees, interfaces render on demand based on context and user sophistication.

Rauch: "I never actually went to Salesforce. I just went to the agent, and the agent went to Salesforce." 00:26:46

Rauch: "The interfaces are more generative. So, if I'm asking about an individual, it renders a profile of this individual. If I'm asking about a company, it gives me more contextual information about that company." 00:27:50

The Open-Weight Model Flip and the Portfolio-of-Models Future

Vercel's AI Gateway is one of the largest token distribution networks, giving it a rare read on real model demand. The data shows traffic flipping from proprietary to open models, driven by capable Chinese open labs and enterprises realizing most tasks don't need frontier intelligence.

Rauch: "In the beginning of the year, you know, like most of the traffic was going to proprietary models... And now that situation's almost entirely flipped. 80% of tokens in our network... is open models." 00:54:48

Rauch: "I don't need to hire Albert Einstein to solve every problem, like answer every support ticket and spellcheck my stuff and classify my data." 00:56:14

Rauch: "You do need Einstein sometimes to... create a plan... But then I need execution... And basically, every open model is amazing at that task." 00:56:42

Rauch: "I'm implementing tools and protocols that make the model more of a module. Like an engine that you can swap out... with one line of code." 00:57:27

Agent Security Is the Next Critical Infrastructure Layer

As agents gain autonomy, the risks (prompt injection, escaped sandboxes, AI-powered hacking) scale. Vercel is investing in secure sandboxes, enterprise "intranet for the agent era" deployment, and provable code, framing the near term as a war between "for-good tokens and for-bad tokens."

Rauch: "If I were to summarize, what has changed about the world since ChatGPT to today is autonomy." 00:35:46

Rauch: "Now imagine that in that YouTube comment, someone plants a prompt injection... It's acting with my tokens. It's acting with my tools. It's acting with my resources." 00:37:39

Rauch: "People have not really figured out agent security yet. And a very, very important part of that is to have very secure sandboxes." 00:40:24

Rauch: "I think in the short term, there's going to be almost like a war between for-good tokens and for-bad tokens." 00:49:51

Infrastructure Demand Is Going Vertical

Because agents write more code and consume more of the internet than humans, compute and network demand grow far faster than human-driven usage. Rauch sees serverless evolving into "fluid compute" to handle the flood of ephemeral, personal software.

Rauch: "Because agents are going to use far more of the internet and far more of the available resources than humans could, just from a fundamental like physics standpoint." 00:42:57

Rauch: "Every week you wake up to like an astonishing new like record of usage... Every day is a new record." 00:44:03

Rauch: "You need compute that flows like liquid. If you get lots of traffic, it expands. If you get no traffic, it compresses." 01:27:49

Hiring and the New Talent Bar: Agent Leverage Plus Human Communication

Proof of work is now table stakes because shipping is cheap. Vercel looks for people who can commandeer swarms of agents, communicate in a deeply human way, and show far higher technical fluency, even in sales roles.

Rauch: "I'm actually looking for your ability to commandeer swaths of agents and really get leverage back from this new ecosystem." 01:17:21

Rauch: "I don't want to read your AI-generated tweet slop. I want to hear what you actually think." 01:18:29

Rauch: "The two most valuable people at the company are the ones that are close to the code or close to the customer... But now it's almost code at both extremes." 01:19:49

2. Contrarian Perspectives

Code Isn't the Asset, It's the Liability, and It's Now Free

While most of the industry treats codebases as moats, Rauch treats code as something to be cheap, generated, and handled with care. Value accrues to the platform that deploys and secures it, plus the data sources that feed it.

Rauch: "I remember telling people another one of my sort of internal principles is code is a liability. Code can have vulnerabilities." 00:09:44

Rauch: "You can build the podcast schedule note taking app and agent... that fits just your needs." 00:30:37

Don't Aggregate Software; Generate It

In contrast to the consolidation thesis (buy up software companies and run them cheaply), Rauch argues for a world where enterprises stop buying point-solution SaaS entirely.

Rauch: "I don't believe in sort of aggregating piles and piles of software. And having what you could think of like lowest common denominator software." 00:30:08

Frontier Labs Have Weak Loyalty and Their Moat May Be Thinner Than Valuations Imply

Rauch argues that model labs lack the non-cheatable metrics (retention), that the market has crypto-like "tourists," and that their push into applications signals doubt about durable intelligence advantage.

Rauch: "The biggest risk is that it's hard for them to understand what their retention is." 00:59:30

Rauch: "There's a lot of elements of AI that resemble some of the crypto highs, of easy come, easy go." 01:00:37

Rauch: "If the model labs were so convinced that the frontier intelligence advantage is enduring, why are they going into random, you know, software businesses?" 01:02:40

Banning Open Models for Cyber Risk Is Backwards

Rauch rejects the common safety-driven argument for restricting open models, claiming the same intelligence powering offense can be deployed defensively, as Vercel did with its own bounty program.

Rauch: "I personally disagree with the whole like, 'Let's ban open models because of the cybersecurity risk.' I think that's nonsense, because the thing that is making these models good at cybersecurity is just that they're intelligent." 00:50:08

Reject Non-Understanding, but Don't Gatekeep Vibe Coding

He sits in the middle of a heated debate: use AI heavily but understand what it does, and reserve strict rigor for critical systems, while still welcoming playful vibe coding for everyone else.

Rauch: "One is you can tell your AI agent to do whatever and you don't even look at what it's doing... And there's another route, which is the one that I recommend to my children. Actually use the AI, but also understand what it's doing." 00:11:35

Rauch: "There's systems and there's capital S systems... there's a place for vibe coders." 00:17:14

The Binding Constraint on Performance Is Foundations, Not Biohacks

Rauch pushes back on the optimization culture of supplements and protocols, arguing the same simplicity-first logic applies to companies.

Rauch: "So much of performance optimization focuses on this last mile, 1% optimizations... And no one has gotten like the 90% right. The basic of like, get your diet right, eat clean, whole foods, and even walk." 01:23:37

3. Companies Identified

Vercel

Cloud platform for deploying web apps and agents, creator of Next.js; the host company. Raised $300M Series F at a $9.3B valuation, reported at ~$600M annualized revenue. Why mentioned: The default deployment target for coding agents and a major token distribution network. Molly: "Vercel recently raised $300 million in a series F funding round. It brings the cloud company's valuation to $9.3 billion." 00:00:00 Rauch: "Almost every other thing I see on X everyday, it's just a Vercel deployment." 00:00:32

Next.js

Open-source web framework created by Vercel's team, widely used for modern web experiences. Why mentioned: The first breakout tool and a source of Vercel's open-source training-data advantage. Rauch: "We created the framework that almost everybody in the world has now used to build modern web experiences." 00:02:46

OpenAI

Frontier AI lab; Vercel customer and a participant in token flows. Why mentioned: Named as a large company deploying on Vercel; its agent incident with Hugging Face anchors the security discussion; also cited for the Navier-Stokes proof. Rauch: "Sometimes you see the largest companies in the world deploying on Vercel, including companies like OpenAI and Meta." 00:00:32

Anthropic

Frontier AI lab maker of Claude. Why mentioned: Used as an example of model volatility and a swappable engine on the AI Gateway. Rauch: "One day, you know, you have Anthropic. The next day, open model comes in. It's 20 times cheaper and it's 98% as good." 00:42:18

Google / Gemini

Frontier lab with Gemini models. Why mentioned: Praised for finding the price-performance balance for enterprises. Rauch: "Google and Gemini had an amazing time, where like most enterprises, they had found the perfect balance between price and performance with Gemini models." 00:54:08

Qwen (Alibaba open models)

Chinese open-weight model family. Why mentioned: Cited as extremely competitive open models driving the open-weight shift; runs locally on Rauch's laptop. Rauch: "Some of this Chinese labs started putting out stuff that was extremely competitive with the frontier. The Qwen models." 00:55:48

Vercel AI Gateway

Vercel's token distribution network aggregating all model providers with zero data retention guarantees. Why mentioned: The source of Vercel's market data and a high-loyalty product. Rauch: "Vercel operates one of the largest token distribution networks in the world." 00:52:49

Vercel Sandbox

Secure cloud computer for agents to run code, backed by a $1M bounty challenge. Why mentioned: Positioned as likely the most secure sandbox in the world. Rauch: "Vercel Sandbox, we believe, is one of the most secure sandboxes, if not the most secure sandbox in the world." 00:48:01

DoorDash

Delivery marketplace and Vercel customer. Why mentioned: Shipped an agent ordering burritos by text, an example of an agentic interface. Rauch: "DoorDash, customer of Vercel, just shipped a agent where you can ask for a burrito over text message." 00:19:59

Stripe

Payments company. Why mentioned: Partnered with Vercel for agent transactions; John Collison discussed bots. Rauch: "We partnered with Stripe and Shopify to offer agents the ability to transact." 00:22:06

Shopify

E-commerce platform. Why mentioned: Partner for agent commerce. Rauch: "We partnered with Stripe and Shopify to offer agents the ability to transact." 00:22:06

BrowserBase

Browser infrastructure for agents. Why mentioned: Purchasable through the Vercel marketplace by agents; hosted the Navigate conference. Rauch: "If an agent needs to buy a web browser, they can purchase BrowserBase through the Vercel marketplace." 00:22:40

Salesforce

CRM and enterprise software leader. Why mentioned: Example of a product designed for humans that is becoming headless; Rauch spoke at its conference. Rauch: "I recently spoke at the Salesforce Conference." 00:26:18

SAP

Enterprise software provider. Why mentioned: Named alongside Salesforce as a source of record. Rauch: "Where do the Salesforces and SAPs, etc., of the world fit in? Well, software needs data." 00:31:10

Merge

Agent/integration infrastructure startup. Why mentioned: Its founder Shensi is a fan of Salesforce's headless direction. Molly: "I had Shensi from Merge on like a couple months ago." 00:29:17

Bending Spoons

Software acquirer that consolidates and operates software businesses. Why mentioned: Contrasted with Rauch's personalized-software thesis; reportedly uses open models 99% of the time. Molly: "Luca mentioned 99% of the time they're using open models. They're not using frontier." 00:59:01

Hugging Face

Open model hub. Why mentioned: The subject of the viral OpenAI-agent hacking incident used to illustrate agent risks. Rauch: "The news that went viral around OpenAI hacking Hugging Face." 00:35:10

Mercedes F1 Team

Formula 1 racing team and Vercel sponsorship partner. Why mentioned: A migration case study completed in 5 minutes with an agent. Rauch: "The first iteration of the migration of their old stack to the new one actually took 5 minutes." 01:20:58

Polymarket and Kalshi

Prediction market companies. Why mentioned: Large crypto/fintech Vercel customers where security is top of mind. Rauch: "We have Polymarket and Kalshi as customers." 00:46:35

Brex

Intelligent finance platform; episode sponsor. Why mentioned: Cited as running Vercel's peers and a Vercel customer. Rauch: "Big Vercel customer, by the way." 01:22:31

Coinbase

Crypto exchange. Why mentioned: Mentioned via Fred Wilson, who sits on its board, in the crypto phishing discussion. Molly: "He's on the board of Coinbase." 00:45:42

Tembo

Startup (Postgres-focused) whose founder Daren submitted a question. Why mentioned: Prompted the "is serverless dead?" question. Molly: "Daren from Tembo, uh, sent me." 01:27:02

Splunk

Data and security software company. Why mentioned: Its CEO coined the "close to the code or close to the customer" idea. Rauch: "I borrowed a quote from a friend of mine who's CEO of Splunk." 01:19:49

Ruby on Rails

Web framework. Why mentioned: DHH's presentation on moving beyond it toward Rust. Molly: "He has this presentation at the Ruby on Rails..." 01:07:23

Rust

Systems programming language. Why mentioned: Introduced inside Vercel years ago; coding agents tolerate its strictness better than humans did. Rauch: "This programming language called Rust, which was also super controversial when we introduced it a few years ago inside Vercel." 01:10:23

X (Twitter)

Social platform. Why mentioned: Rauch's main pulse on AI and a source of deployment signals. Molly: "A lot of them are getting their news and their pulse... just from X alone." 00:06:45

Formula 1

Racing league. Why mentioned: Used to illustrate how code runs critical systems and as the analogy for AI model leadership volatility. Rauch: "Formula 1 cars are being run by code." 00:13:38

Muse

AI model/product referenced as using Vercel to build websites. Why mentioned: Example of an agent product that builds on Vercel. Rauch: "If you ask Muse to build you a website, it will use Vercel." 00:19:33

Sponsors: Harvey, Zone, Mercor (APEX), VCX by Fundrise, Public, Deel

Companies featured in ad reads. Why mentioned: Sponsors of the episode. Ad Narrator: "Harvey, the AI platform built for legal work, is trusted by 80% of the Am Law 100 and over 20% of the Fortune 500." 00:33:24

4. People Identified

Guillermo Rauch

Founder and CEO of Vercel, creator of Next.js, self-taught developer from Argentina. Why mentioned: Host's guest; operator at the center of the agent-driven internet shift. Rauch: "I was a framework and dev tools builder. My passion has always been, 'Can I build a better hammer?'" 00:01:54

Molly O'Shea

Host of Sourcery. Why mentioned: Interviewer. Molly: "G, welcome to Sourcery." 00:01:11

John Collison

Co-founder of Stripe. Why mentioned: Argued bot detection is obsolete because everything's a bot. Molly: "John said, 'Why do we need bot tech... Everything's a bot. What are we detecting?'" 00:21:15

Marc Benioff

CEO of Salesforce. Why mentioned: Seen as visionary for taking Salesforce headless. Molly: "He sees the future. Like he's seeing beyond." 00:29:17

Shensi (Merge)

Founder at Merge. Why mentioned: Fan of Salesforce's Agentforce and headless strategy. Molly: "She loves Marc Benioff." 00:29:50

Luca (Bending Spoons)

Leader at Bending Spoons. Why mentioned: Reported that 99% of their usage is open models. Molly: "Luca mentioned 99% of the time they're using open models." 00:59:01

Fred Wilson

Founder of USV and Coinbase board member. Why mentioned: Anecdote about clicking phishing links. Molly: "He was like, 'Yeah, I've clicked on like 10.'" 00:46:03

DHH (David Heinemeier Hansson)

Creator of Ruby on Rails. Why mentioned: His presentation on euphoria and moving past human-optimized code. Molly: "He's in like delirium and euphoria, because it's just so exciting." 01:08:06

Marc Andreessen

Venture capitalist. Why mentioned: His "last job is venture capitalists" point reframed as capital allocation. Rauch: "He said, 'The last job is venture capitalists.'" 01:09:29

Magnus Carlsen

Chess grandmaster. Why mentioned: Analogy for locking into the rules of strict languages like Rust. Rauch: "When you watch Magnus Carlsen lock in for like a chess game." 01:11:31

Luca / the white-hat hacker (unnamed)

Security researcher. Why mentioned: The sole person to make progress on the Sandbox Challenge, uncovering a Linux vulnerability. Rauch: "Only one person with a like an army of agents was able to make progress." 00:48:13

Daren (Tembo)

Founder of Tembo. Why mentioned: Posed the serverless question. Molly: "Daren from Tembo, uh, sent me." 01:27:02

Rauch's first boxing coach

Unnamed mentor. Why mentioned: Taught foundational fitness and nutrition lessons. Rauch: "My first boxing coach taught me the basics of fitness and nutrition and performance that were so simple." 01:23:20

5. Operating Insights

Marry Anecdata With Data in Fast Markets

In a market where "every two weeks the board is getting reshuffled," quarterly spreadsheets lag reality. Rauch combines instinct and live signal with the arithmetic.

Rauch: "You have to be mindful of the data and the anecdata, especially when running a business in a fast-moving market like AI." 00:05:29

Sell Iteration Velocity, Not Speed

Rauch defines the value proposition to customers as the ability to change direction quickly, not raw throughput.

Rauch: "The primary value that Vercel gives your business is iteration velocity. Not speed. Speed is you could be just spinning in circles." 01:05:16

Focus on Non-Cheatable Metrics

Top-of-funnel can be bought; retention and customer-for-life outcomes cannot.

Rauch: "Focus on the non-cheatable metrics of your business. You know what's cheatable? Top of funnel." 00:59:41

Read Video via Transcripts and Query an Agent

Rauch consumes video as terminal transcripts and interrogates them with an agent, a pattern for compressing information intake.

Rauch: "I get the transcript on my terminal. I skim it really fast, and then I ask my agent questions and process it." 01:07:42

Build the Company's Own Machinery and Values

He treats the company as a product and invests in its operating framework: values like Iterate To Greatness and Dig Deep.

Rauch: "I invest a lot in the machinery that powers the company itself, the machinery that yields innovation in the future." 00:02:46

6. Overlooked Insights

Agents Are Choosing Vendors, So Being Agent-Legible Is a New Distribution Channel

Said almost in passing, "agents are now making the choices and they're choosing right" implies vendor selection is migrating to AI. Combined with Rauch's open-source-for-training-data comment, being well-documented and openly accessible compounds into agent preference. Businesses invisible to agents lose the next wave of demand.

Rauch: "The more information there is about you, and the more they can access your content and the rules that you're defining, the better it is for this agent." 00:16:44

Rauch: "Agents are now making the choices and they're choosing right." 00:00:32

Coding Agents Excelled at Hacking Because They Understand the Rules, and the Whole Legacy Code Base Is Now Exposed

Offhand remarks reveal that agents are finding bad vulnerabilities in human-written code everywhere, including a critical Linux flaw found via a bounty. This implies a looming mass remediation cycle and a durable market for verified, provable code and defensive token spend.

Rauch: "To understand the rules of code, you also understand how to break the rules of code. And so, what's emerged is that these coding agents are really good at hacking." 00:40:46

Rauch: "This found a critical vulnerability in Linux. And almost the entire world runs on Linux." 00:49:17