Resend CEO: Why Great Startups Enter Bad Markets

Today I interviewed Zeno Rocha, co-founder and CEO of Resend, one of the most used developer platforms for sending email, with more than four million users and 100,000 paying customers.

The question everyone asked him at the start was some version of the same one: why would you build an email company? The market was crowded and consolidated, every good product in it had been acquired, and none of them had become huge on their own.

He went in anyway, without knowing anything about email. Then AI arrived and turned the boring market into one of the fastest growing ones in dev tools, with agents as his fastest growing customers.

Structure and Key Highlights

The Two Gifts that Changed the Trajectory

• His father pushed English on him early, which he says is uncommon in Brazil
• The reasoning was simple: the knowledge available in English has no comparison with what exists translated
• They grew up poor, and his father treated education as the way out
• His grandfather gave him a computer around age ten. He was never a programmer, he had been a pilot at one point, and yet his bookshelf had database books and a C# book
• His own read: without those two things he would not be here

Tweeting in English to Ten Followers

• He started tweeting in English around 2008, while still living in Brazil, and his friends mocked him for it
• He came up through Flash. When Steve Jobs wrote that Flash would not run on iPhones, the whole market had to move
• Everything he needed to learn HTML5 was in English, and his first open source project and English website made the world feel bigger than his neighborhood
• He moved to the US around 2012. His framing is that you plant seeds and they take two or five years

A Complicated Love for Open Source

• He fell in love with it at the exact moment freelancing was paying him more than his day job, while open source paid nothing
• What it gave back was different: people used what he built and challenged his ideas with use cases he had not considered
• Around 2015 he was the 23rd most active user on GitHub, when the platform already had a couple of million users

What AI Did to Open Source

• Every all-hands they still review issues opened and PRs closed, and highlight contributors
• The sad part is what those contributions look like now: someone fires up ten agents and sends PRs that do not solve a real problem
• He expects less of the community feeling and more caution, because with the security breaches of the last year, adding a dependency is genuinely scary
• And since code is cheap to generate, the calculus changed: bring it in-house and maintain it yourself
• Resend has an open source component but is not an open source company. The main product will not be open sourced

A CEO Who Still Answers Tickets on Weekends

• Coding turned into a creative outlet for him, one of several ways to express an idea, alongside a document or a presentation
• He still reads the PRs and sends one now and then, even if he is not the one reviewing
• Everything at Resend lives in code, including the marketing site and the blog
• Every new hire, technical or not, writes a welcome post on day one. That means installing a terminal and sending the first PR of their life
• He does support on weekends on purpose, because the ticket queue is what keeps him close to what customers actually feel

Why Email Looked Like a Terrible Idea

• SendGrid, Mailgun and Postmark all ended up acquired. Good outcomes, but not the independent multibillion dollar companies
• He loved SendGrid in 2016, watched Twilio buy it for $2 billion, and then watched the product decay
• His read on tech: every ten years or so the stack gets refreshed. Auth0 gave way to WorkOS and Clerk, Heroku to Vercel
• They wanted to be that for communications, and they came in naive: they knew how to build for developers, not email
• His own line: if he knew then everything he knows about the market now, maybe he would not have started the company

The Part of the Business He Never Talks About Publicly

• The market is nasty in two directions: the competition, and the volume of bad actors trying to send phishing through you
• He has personally spent a lot of time on abuse, and it is the piece of the job he had never discussed in public

Care as the Wedge

• If a bomb dropped and he had to start over, he would pick an insanely big market, and specifically one with many competitors where nobody seems to care
• The bar he wanted to import was the care he saw in products like Stripe and Vercel
• His argument is that care is legible from the outside: when you give a shit about a problem, people notice and reward you with attention
• So the product carries it in small details, repeated constantly

Watching the Wave from the Beach

• Around January 2025 he watched Lovable, Bolt and v0 go from zero to $100 million in ARR, and describes it as sitting on the beach watching surfers
• By January 2026, with Claude Code and Opus, Resend was surfing it
• That changed the trajectory of the business, and he says 2026 is still insane for them

The Definition of a Developer Changed

• They see around 400,000 new signups a month, and he estimates at least 20% are non-technical people testing whether the product is for them
• His example: an email from a 51-year-old who does not know how to code, was pointed to Resend by an AI, was put off by the dark mode hacker vibe and by not knowing what an API key is, and still sent his first email
• Those users pushed the onboarding from five steps to four, then three, and now two
• When people started asking to just tell Lovable to send an email, they built integrations, OAuth and MCP servers
• He welcomes the pressure: lowering the bar of entry makes the product more global

Low Floor, High Ceiling

• During YC they had no way to charge, so he begged people on calls to try the product and closed by sending a Stripe payment link by hand
• The next layer was Stripe Elements, embedded in the app, still without building the full integration
• At a thousand paying customers they built the deep integration with their own look and feel
• That is his model of a good product: one problem solved at several abstraction layers, from no-code to full API

The Company as a Product

• His framing: if my company were a product, what would its pricing page be, its about page, its homepage
• After six months of hitting walls on product decisions with no clear answer, they sat down and wrote the philosophy of the product
• Two of those principles: docs are part of the product, and everything is in the details
• Later it became a full handbook, published in public, applying the build-in-public idea to a company instead of a personal brand
• The throughline is care. The internal phrase is that they will out-care everyone in the industry

Agent Experience Is Not Developer Experience

• Agents are an extension of the user now, and a lot of what makes developers happy means nothing to them
• Dark mode is the clearest case: agents cannot see it
• Their first landing page made you use the keyboard to join the waitlist, as a wink to developers who care about shortcuts. Agents do not care
• People still tweet every week about the animated icons in the dashboard. Agents dislike the rotating Rubik’s cube on the landing page because it is more bytes to load
• What did carry over: REST API design, SDKs and the OpenAPI spec went from nice-to-have to must-have
• The CLI he had filed away as a fun hack week project turned out to be a real need, because personal agents need one

Getting Recommended by a Model Is Now a Channel

• A third-party study wrote more than 2,000 prompts across models to see which tools they recommend when you say something like I need feature flags or I want to send an email
• Resend came up 67% of the time on Opus, and ranked first again in a later study on Fable
• So he studies the patterns, keeps an eye on llms.txt, and now ships markdown pages for everything

The Pricing Slider an Agent Could Not Read

• He asked his own agent how Resend’s pricing compared to competitors and it answered that they were the most expensive, which is false
• The cause was mundane: their pricing page uses a JavaScript slider
• Agents scan a page, and even with computer use they cannot drag a widget
• The fix was a markdown version of the pricing page served to agents

Seven Agents with Their Own Inboxes

• He buys the framing of an agent as a new kind of user, and runs it that way personally: a trainer he calls Arnold that counts calories from photos of his food and holds three years of blood tests, plus others for writing and for relationships
• One of them is a personal COO he calls Tim, loaded with the Resend handbook and instructed to push back and tell him hard truths instead of pleasing him
• At company scale there are seven agents: marketing, ops, trust and safety, sales and more
• He does their onboarding himself and says it feels like onboarding a human: each has its own email inbox, GitHub account, 1Password vault, its own machine and a seat on Slack with its own avatar
• They pick up the personality of their team. The marketing one jokes around, the infrastructure one does not
• Access is scoped deliberately, like read-only Datadog so an agent can jump into an incident channel, triage it, and answer afterwards which customers were affected

80% Is Easy, 110% Is Harder Than Ever

• His compression of the paradox: it has never been easier to reach the 80% solution and never harder to reach the 110%
• What he pushes internally: it works, great, that is the beginning. Everyone can make it work, so now the job starts
• He welcomes the explosion of dev tools even knowing many of them will not exist in a year
• He does not buy the full SaaS apocalypse narrative, but admits he replaced a subscription with a single chat window
• Infrastructure keeps growing for a simple reason: the more software gets built, the more those foundations are needed

What Kills a SaaS Now

• The ones that die will be those that do not adapt to be surfaced when someone asks an agent, which he calls the new Google
• Data locked behind a signup wall becomes a liability rather than a moat
• As a consumer he is blunt: if a product has no MCP server or connector, he will not use it
• His own example: Arnold pulls his exercise from Strava to subtract calories, and if Strava did not expose that, he would move to a service that does

The Debate They Have Not Resolved

• Stay narrow and be absolutely loved by a small group, or go broader and risk diluting the product
• The practical version of the question is mundane: how does the sidebar look when there are four SKUs to click and buy
• Launch weeks create the other risk. The new shiny thing arrives and the previous product stops evolving
• His own phrase for it, from a blog post: cheap to launch, expensive to maintain

San Francisco, Anxiety and Clarity

• The team is fully remote, which gives him the freedom to be elsewhere, but he says the energy in SF cannot be matched and his people are there
• It was essential for the first years of Resend. As the company became global, being closer to the rest of the team also became important
• He describes constant anxiety and terrible sleep, and says that when agents arrived it became the best and the worst time of his life
• The reason is precise: the moment you do not have an agent running, you feel you are wasting your time
• VC dinners and constant comparison push you to decide from fear and from missing out, and produce urgency for the sake of urgency: we have to ship this today, why, because we need to move fast, why
• His conclusion is that the CEO job runs on clarity, and distance is what buys it

Who He Thanks

• Diana, his YC partner, who he says has always believed in him more than he does
• Before the batch even started she asked for his end-of-year goal. They had two users: his best friend, and himself paying with his own Gmail
• He guessed he could get to ten. She said thirty. They closed the thirtieth customer on the last day of the year
• She did it again with an ARR goal he thought impossible, and they hit that one too

Silicon Valley Glossary

People, Companies and References

People Mentioned

• Zeno Rocha: co-founder and CEO of Resend
• Diana: his YC partner, the person he thanks for making him think bigger
• Steve Jobs: the open letter about Flash that pushed him toward HTML5 and English
• Arnold: the name he gave his personal trainer agent
• Tim: his personal COO agent, told to push back instead of pleasing him
• Enzo Cavalie: podcast host

Companies and Organizations

• Resend: email for developers, four million users and 100,000 paying customers
• SendGrid, Mailgun, Postmark: the previous generation of the market, all acquired
• Twilio: bought SendGrid for $2 billion
• Auth0, WorkOS, Clerk: his example of a category being refounded
• Heroku, Vercel: the same pattern in hosting
• Stripe: the product he copied the care from, and the payment links that got him through YC
• Lovable, Bolt: the products he watched go from zero to $100 million in ARR
• Claude Code: what made the wave reach them in January 2026
• GitHub: where he was the 23rd most active user around 2015
• Supabase, Firecrawl: infrastructure companies growing faster than ever
• Datadog, 1Password, Slack: what his agents get accounts and scoped access to
• Strava: the consumer product whose MCP server his own agent calls every night
• YC: where Resend started, with two users

Data and References

• 4 million users and 100,000 paying customers: where Resend is today
• 400,000 signups a month: the current pace, at least a fifth of them non-technical
• 67%: how often models recommended Resend in a third-party study on Opus
• 2,000 prompts: the size of that study
• $2 billion: what Twilio paid for SendGrid
• 5 steps to 2: how far the onboarding was compressed
• 23: his rank as most active GitHub user around 2015
• 7: the agents running inside Resend, each with its own inbox
• 2 users: what they had when his YC partner asked for the year-end goal, one of them himself