Oct 17, 2025 · 1h 13m · 20vc
20VC: The Startup Adding $1M ARR Every Week | Competing Against OpenAI's Codex and Claude Code: Who Wins | Why Gemini is Failing and GPT-5 Is Winning | Do Margins Matter in a World of AI | The Ugly Truth About AI Coding with Zach Lloyd, Warp
0:00 / 0:00
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of 20VC, host Harry Stebbings interviews Warp founder and CEO Zach Lloyd about building a next-generation AI developer terminal, navigating the unit economics of AI SaaS, and competing against Big Tech and frontier model providers.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 36.9% of the talking time here. How this is scored →
Harry as informed peer 5.0 Guest teaching 4.2 Guest disagreement 3.4 Harry pushing back 4.4
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 26:37 Zach rejects VC trope on product moats
Zach forcefully calls bullshit on the prevailing VC opinion that software product quality is not a moat, directly challenging Harry's thesis that Warp users are easily switchable.
Hardest push from Harry ▶ 37:09 Harry calls board advice shitHarry explicitly refuses Andrew Reed's board guidance to balance growth, margin improvement, and enterprise expansion all at once, calling it unrealistic shit advice that ignores necessary startup trade-offs.
Biggest teaching moment ▶ 6:01 Zach educates on early startup rewrites vs Google scaleZach explains from engineering experience why rewriting code early destroys startup speed, reframing Harry's venture capital perspective on technical debt.
Harry holds his own ▶ 8:55 Harry lists Google's asset portfolio against Zach's bear caseHarry demonstrates deep market knowledge by countering Zach's bearishness on Google, citing YouTube, Waymo, and Google's 10% equity stake in SpaceX.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Engineering Lessons from Google Sheets to Startups | 5 | 6 | 4 | 6 | Harry challenges Zach's rule against code rewrites, arguing early rewrites are better than taping technical debt together. Zach reframes the issue around startup velocity, explaining that rewrites before product-market fit are fatal engineering traps. | |
| Critical Evaluation of Google Culture and AI Strategy | 7 | 3 | 3 | 5 | Harry pushes back on Zach's skepticism toward Google's future by listing Google's hidden asset portfolio, including Waymo, YouTube, and its 10% equity stake in SpaceX. Zach acknowledges those assets but maintains that Google's corporate culture is dangerously risk-averse. | |
| Comparing Frontier AI Models for Developer Workflows | 4 | 4 | 2 | 3 | Harry asks for technical comparisons between GPT, Claude, and Gemini, proposing a simple split where OpenAI wins consumer and Anthropic wins enterprise. Zach nuancedly breaks down model behaviors and notes that developer preferences remain open. | |
| Agentic Coding, Vibe Coding, and Developer Impact | 5 | 5 | 3 | 4 | Harry questions whether agentic coding tools actually have genuine product-market fit given mixed productivity studies despite astronomical valuations. Zach agrees that naive vibe coding fails in production and that senior developers are needed to extract real value. | |
| The Economics of AI Products and Margin Realities | 5 | 4 | 4 | 6 | Harry directly probes Warp's margin profile, complaining that VCs only ever hear top-line revenue figures. Zach acknowledges consumer margins are tough due to API costs, turning the table to ask Harry how VCs evaluate fast-growing but margin-negative AI startups. | |
| Moats in Dev Tools vs. Model Providers and Cursor | 6 | 5 | 7 | 7 | Harry bluntly challenges Zach's defensibility, arguing developers are easily switchable to OpenAI or Anthropic. Zach forcefully rejects the VC consensus that software product is not a moat, calling it bullshit when applied to Warp's deeply customized terminal environment. | |
| Navigating the Frothy AI Capital and Deployment Market | 5 | 4 | 2 | 3 | Harry reflects on market froth, citing 200x revenue multiples and circular deals like NVIDIA funding OpenAI. Zach agrees the market is wild but defends the core thesis that AI technology is genuinely transformational. | |
| Sequoia Board Advice on Balancing Growth and Margins | 6 | 5 | 5 | 8 | When Zach shares Andrew Reed's advice to maintain growth while gradually improving margins, Harry rejects the framing and calls it typical unhelpful VC advice that ignores hard trade-offs. Zach defends Sequoia's strategic guidance while acknowledging the difficulty. | |
| Long-Term AI App Margins and Commodity Model Markets | 4 | 5 | 3 | 3 | Harry and Zach explore long-term AI application margins and whether foundation models will commoditize. Zach explains why coding models may reach a 'good enough' threshold where application UX and context retrieval capture most value. | |
| OpenAI Valuations and Warp's Preempted Funding Rounds | 6 | 3 | 2 | 4 | Harry points out that early OpenAI investors haven't seen extraordinary percentage returns relative to venture risk due to extreme dilution. They discuss Sequoia's preempted $50M Series B check into Warp prior to any revenue monetization. | |
| Sequoia's Cheat Codes and the Marc Benioff Connection | 3 | 4 | 2 | 2 | Zach illustrates Sequoia's network value by describing how Andrew Reed got the CrowdStrike president on the phone within hours to fix a block on Warp. Zach also reveals that Marc Benioff is his second cousin. | |
| Why Big Tech Incumbents Lag in AI Dev Tools | 4 | 3 | 4 | 3 | Zach criticizes Big Tech incumbents for slow execution and bureaucratic drag, contrasting Google's slow voice assistant updates with startup speed. Harry echoes the frustration using Apple's keyboard constraints. | |
| Quickfire Round: Granola, Ambitions, and Geopolitics | 5 | 4 | 3 | 3 | In the quickfire round, Harry candidly admits passing on Granola and telling its founder to start a GoFundMe. Zach discusses long-term personal ambitions outside tech, including public service. |
Statements from this episode (29)
Stebbings: Warp raised over $70M from Sequoia, GV, and Marc Benioff
“They've raised over seventy million dollars from the likes of Sequoia, GV, Dylan Field, Elad Gill, and even the legend that is Mark Benioff.”
Google Sheets Had Over 100 Million Users During Code Rewrite
“We rewrote Google Sheets because it's a crazy scale. And so, you know, if you have something that has a hundred million plus users, which is what it had at the time, I think it's now like a billion users.”
Lloyd: Codebase rewrites are a horrible idea that pause startup momentum
“For the audience for this show, which I assume is more founders, like rewriting is like a horrible idea. It's like pausing time.”
Lloyd: Google faces a classic innovator's dilemma in AI
“I think Google's a risk. Like, I would never have said that five, 10 years ago. I find myself using ChatGPT more and more, and so I think they have big innovators dilemma problem.”
Lloyd: OpenAI and Anthropic models beat Google's Gemini in coding benchmarks
“The two leaders are GPT-V and the Claude models, and so we measure them. Like, first of all, it's like, it's not so much gut. It's like we literally We have a whole bunch of benchmarks. We run sweet bench. We run terminal bench. Claude and GPT-V are the two be…”
Zach Lloyd: No AI provider has a strong moat at the API layer
“I don't know that anyone has like a super strong moat at the API layer, just to be totally clear.”
Lloyd: Every coding task at Warp starts with a prompt
“And so every single coding task we do starts with a prompt, which is the big shift in productivity that's happening right now.”
Lloyd: Autocomplete has stronger enterprise PMF than agentic AI
“In the enterprise market, I actually think the thing that has most product market fit is like autocomplete. Which is not the sexiest type of AI. So it's not agentic AI. It's like cursor's original product or even co-pilot.”
Lloyd: Engineering teams will shrink to fewer senior engineers managing agents
“I think we're gonna have less. I think we're gonna have fewer, more senior engineers who are managing the work of a lot of agents.”
Lloyd: Fixed-fee AI SaaS margins worsen as product usage increases
“The more they use it, the more it costs you. So the better your product market fit in a sense of like the product is getting better, the like worse your business is from like that perspective.”
Zach Lloyd: Warp is adding $1M in net new ARR weekly
“We're now adding a million net new ARR every week, or even less. Honestly, recently it's accelerating.”
Lloyd: The VC trope that product quality is not a moat is false
“I feel like there's like a VC trope right now around like, well, product's not a moat. I think that that's like kind of bullshit, especially when everyone else in the market has the same product.”
Lloyd: Warp has 700,000 users, outpacing Claude Code and OpenAI Codex
“We have way, way more users than Cloud Code, for instance, or Codex. We have 700,000 users because we've been doing this for Five years.”
Lloyd: Prompt-driven development, not autocomplete, is the future of software
“I don't think autocomplete is the future. So I don't feel like I'm like seeding competitive ground here to Cursor. The future is more like developed by prompt.”
Lloyd: Demand for senior engineers is increasing despite AI advances
“Which again, I think reinforces the like, the importance of senior engineers is not going down. It's never been more competitive for us to hire someone good.”
Lloyd: AI deployment will lag technology capabilities over a ten-year horizon
“I think it's more like a 10. I think, like, again, the technology's gonna run way ahead of the deployment of it, just because the deployment of it takes a long time.”
Stebbings: Startups cannot simultaneously fix margins, maintain growth, and move upmarket
“There is a world of trade-offs, which is like, I cannot sustain growth, focus on margins and improve them, move into enterprise, or sustain it and make it super happy. Oh, and continue to focus on beautiful product quality all at once.”
Lloyd: Anthropic API Margins Are 60%, OpenAI Margins Are 50%
“Anthropic has a big margin on their API business right now. What I've heard is 60% margin on Anthropic's API. I've heard 50% on OpenAI.”
Lloyd: Open-Source Models Will Not Win at the AI Frontier
“I don't think open either, if I'm being honest. The reason I don't think open, like, is that it's just so damn expensive to build these things, and so what's the economic incentive?”
Lloyd: Warp will reach 1M active developer users this year
“I think we'll hit a million active users, active developers this year, and we're, and now obviously we're growing our revenue super duper fast”
Lloyd: Sequoia connected Warp to CrowdStrike's president within hours of an issue
“And so I reached out to Andrew. I was like, can you help me here? And then that day I was on the phone with the president of CrowdStrike.”
Zach Lloyd: Microsoft screwed up GitHub Copilot
“I guess, but they, like, I think that they kind of screwed up Copilot.”
Zach Lloyd: Google is bad at building developer tools
“Google is not great at developer stuff, just period from my time there.”
Lloyd: Big Tech incumbents are failing to build great AI dev tools
“I don't think any of those companies are doing a great job as I insult all of our potential acquirers, but like, I, it's just like, I just don't like, I think that It's like, there's a big company thing that's a real problem.”
Stebbings: I passed on Granola and told the founder to launch a GoFundMe
“Not only did we not invest, we said you should set up a GoFundMe page.”