Hiring Guide

Working at Replo

Hi! My name is Yuxin (pronounced yoo-shin) and I’m one of the co-founders of Replo. If you’re reading this, we’re probably chatting about working together.

This doc answers some basic questions about us: how we work, what we're building, and our strengths and weaknesses as a team.

I’m a big believer that recruiting is a two-way street. This doc is not meant to sell you on Replo. It is meant to help you aggressively decide whether Replo is a good fit. By the end, you should be excited about working with us. If not, please DM me and kindly decline continuing the process (I swear I don’t take this personally).

I don’t want to waste your time or ours on a role you feel 🤷 about. Life is too short to do work you don’t enjoy.

Computers that run your business

In a few words, Replo builds agents that run your business.

We want everyone who sells online to have a world-class GTM team in their pocket. Our agents build the store, manage inventory, run marketing, and watch the numbers. Here are a few things they’re helping our customers do today:

  1. Build stores for three separate Super Bowl commercials.
  2. Analyze billions of data points across dozens of platforms every morning, then design and execute multi-week marketing strategies for companies with millions in advertising spend.
  3. Coordinate with suppliers around the world to launch custom merchandise for YouTube creators and artists.
  4. Manage product launches across storefronts, email, paid media, and customer support.

Sell anything on the Internet. Replo runs the rest.

Why now

“The best entrepreneurs know this: every great business is built around a secret that’s hidden from the outside. A great company is a conspiracy to change the world; when you share your secret, the recipient becomes a fellow conspirator.” — Peter Thiel

Every company should answer three questions: Why now? Why us? Why haven’t the incumbents won?

More businesses will be started in the next 10 years than in the last 100

“People pay for information, but people mostly pay for work. Talking to a chatbot and getting an answer is super great. But getting work done? I’ll pay for that.” — Jensen Huang

Intelligence is becoming cheap and widely available. Agents represent a new computing model — software that can reason, execute work, and operate in parallel on a massive scale.

The old barriers to starting a company — capital, labor, and specialized knowledge — are falling away. More people can build what they imagine, faster. We think that will create a generational renaissance in entrepreneurship and fundamentally empower individuals to bring new products and services to the world.

The limiting factor today is labor, not tools

Software for selling stuff online is already easy to use. Amazon, Etsy, and Shopify exist at the fingertips of millions of people. Launching a website, opening a bank account, and accepting payments can now be done in minutes from your phone.

Yet the number of businesses that succeed remains low.

Starting and growing a business still takes enormous amounts of time, money, and specialized knowledge. Not only are there a million things to do, but each one requires expertise and experience. Launching a website feels monumental. In practice, it is only the beginning. What follows are late nights spent designing the website, making ad creative, writing copy, managing inventory, packing boxes, closing deals, answering customers, and doing the accounting.

The last Internet era built tools. Shopify helped you launch. Klaviyo helped you reach customers. Stripe helped you charge them.

The AI era is different. You don’t want tools; you want outcomes. Not a website, but sales. Not another email tool, but customers who come back.

Replo automates this across tens of thousands of businesses. Our agents monitor performance, identify what needs attention, and execute coordinated work across your storefront, marketing, inventory, and P&Ls. They don’t just tell you what to do. They do the work for you.

Models change, but platforms endure

Software is getting cheaper to build. Features are getting cheaper to copy. The products that endure are the ones where data, workflows, and institutional memory compound over time.

Replo participates in every transaction on our platform. Our revenue is tied to the value merchants create, rather than the number of tokens consumed. When we help merchants create demand, convert customers, and process sales, we share in that value and reinvest it into doing more work for them.

That gives us fundamentally different unit economics. AI companies that mark up inference make more money when their customers consume more tokens. We make more money when our merchants make more money. Our bet is that by maximizing the value created by every dollar of inference, every successful transaction gives us more capacity to make the system better.

Whether we are right about what the next few years will bring remains to be seen, but we hope you will join us along the way.

Our team

Noah and I have been friends since our freshman year at Berkeley. During our undergraduate years, we designed and built a course scheduling system to replace Berkeley’s age-old telephone-based system, Telebears. We ended up selling it to the university when we graduated in 2015.

After college, I joined Uber, where I became a senior staff engineer and one of the youngest engineering leaders at the company. I helped lead the company through its 2019 IPO and launched new financial and commerce products around the world, including Uber Eats, Uber Delivery, and microlending for our partners. While at Uber, I also built a substantial ecommerce business with my roommate at the time, which first brought me into the direct-to-consumer space.

Noah was previously a principal engineer at PlanGrid (YC W12), where he led iOS development before the company was acquired by Autodesk in 2020.

Today, we have no middle managers, and most people work directly with Noah or me. Engineers hop on customer calls, weigh in on product, propose ideas for GTM, and work across whatever parts of the company they think will move us forward. Someone on our team described it as “founder bootcamp,” which I think is pretty accurate.

About the company

We have just under $8M in ARR, $4.2M in cash on hand, and are profitable. To date, we’ve raised $8.4M in seed funding from General Catalyst, Y Combinator, Figma, and Vercel. We have 16 employees, 13 of whom are engineers.

Most of the team is based in San Francisco, and we go into the office 4–5 days a week. Our office is at 800 Market St, where we host poker, board game, and video game nights. It’s an awesome space that we’ve really made our home — please come by and visit if you can!

What type of team are we trying to build?

The best analogy I can come up with is:

It’s junior year at Berkeley. It’s 6 p.m., and the class project is due at midnight.

You’ve assembled your best friends in the basement of the engineering building. The lights flicker. The hum of an ancient mainframe buzzes behind the walls. None of you have gone to lecture the whole semester, but you have enormous trust in your friends — everyone is smart, capable, and on the same wavelength.

You split up who’s gonna do what, and everyone gets to work on their individual tasks. You’re stressed, but you know everything is gonna work out.

At 11:59 p.m., you press submit on the project, and a sudden wave of euphoria passes over you. Amid the chaos and panic, the autograder results come back, and you realize you and your friends are unbeatable.

Small teams win. They’re also much more fun. My job is simple: find exceptional people and give them hard problems.

Reasons why Replo is a great fit

You wanna win (in an almost insane way)

“We learned quickly that the most important predictor of success is determination. At first we thought it might be intelligence.” — Paul Graham

If someone tells you that you can’t do something, you wanna get out there, do it, and prove them wrong. Some people join startups because “working at a startup” sounds like a fun lifestyle. That is a bad reason. You should join because you want to build something GREAT that’s never been done before.

You want actual ownership

“Artists aren’t merely capable of shipping. They insist on it. So if you don’t let people ship, you won’t have any artists.” — Paul Graham

Everyone wants ownership, but usually that means a PM writes a ticket and takes the call from the customer while you decide how to implement the solution. That’s not what we mean. If you find an important enough problem, we expect you to go solve it end-to-end — from product to design to engineering.

You are AI-pilled in an extreme way

Everyone writes code here. Engineers ship features. Marketing builds prototypes. Everyone uses Cursor and Claude Code. If AI is not changing how you work every month, you’re behind.

You have a strong sense of urgency

If something can be done today, you’ll do it today. If an agent can work overnight, you kick it off before bed and leave your laptop plugged in.

You’re scrappy

Scrappy means being relentlessly resourceful. Being relentless isn’t enough — the problems you encounter at a startup are novel, so brute force alone won’t get you through them. You keep the goal fixed while changing the plan as often as necessary. If someone tells you something is gonna need X, Y, and Z, you figure out how to get it done with 10% of the resources and 10x the speed. You don’t wait for the perfect tools, more headcount, or someone’s permission. You keep trying new things until you make it work.

You wanna build a culture of builders

You wanna work with people who are thoughtful, creative, and extremely good at what they do. We argue, debate, and disagree vehemently, but at the end of the day, everyone is working toward the same goal: building the best product imaginable. The hackathons, workout classes, and ski trips aren’t manufactured company culture — they happen naturally because we genuinely enjoy spending time together.

You don’t have time for bullshit

We don’t say fake things to spare each other’s feelings. Don’t be an asshole. Be nice to each other, but don’t sugarcoat hard problems and issues. Celebrate the wins and learn from your losses.

Transparency matters to us, and we want to build a tight-knit team. If you didn’t get to something, just say you didn’t have time. If you need help, ask for help. We’re here to win as a team, and if we don’t keep communicating this way, it’s not gonna be fun for anyone.

Reasons why Replo is probably a bad fit

You’re not technical

If you’re not able to write code, you’re probably not a good fit. We exclusively hire people with strong technical backgrounds and a quantitative mindset, regardless of title. You don’t need a computer science degree. You do need to use LLMs well, spot workflows that can be automated, make decisions with data, and reason from first principles.

You’re interviewing companies big and small

A version of this is that all your friends are working at startups, so you thought, “I might as well give it a go.” Paul Graham has a great article on why you shouldn’t start a startup just because it feels like the thing ambitious people are supposed to do. Startups are all-consuming, lack structure and mentorship, and constantly force you to solve problems you’re not prepared for. You should consciously decide you want to join one before you start interviewing.

Working here will improve your career opportunities down the road because you’ll need to do a bit of everything. It’ll build resilience, motivation, and a can-do attitude that none of your peers will get from working on a button at Meta for two years. There’s a reason recruiters at big tech companies always try to convince you that the team you’re joining is a “startup within a big company.”

You want to be a manager in your next role

You can grow into management here. But you need to join as an individual contributor. Hiring should not be the de facto solution to any problem, and coming in expecting that you’ll be managing the actual makers is a recipe for getting fired. Also read The One Thing CEOs Should Delay as Long as Possible.

You like the “strategy” but not the small details

The number of times a candidate says, “I wanna work on strategy,” during our conversation is inversely proportional to the probability of getting the role. Ideas are cheap. We expect everyone to do strategy, ask AI tools why their ideas won’t work, iterate, and implement those ideas from 0 → 1.

This video is an instructive example of people who only wanna do strategy and don’t actually do the work. If you own a project and hear yourself saying, “I’m not an engineer,” “I’m not in sales,” or “We need to hire someone for that,” Replo is probably not the right fit. If you don’t know how to do it, figure it out.

If Noah and I had waited to hire a recruiter before we hired engineers, we’d never have a team. If we had waited to hire a designer, we’d never have a product.

Things we believe

Here’s a set of contrarian things we believe. They’re more tactical than cultural values, but each reflects an important truth about building companies that we think most people still get wrong.

We do things that don’t scale

We talk to customers, onboard them ourselves, and obsess over their experience — then build everything we learn back into the product. We would rather have a small number of customers who love us than a large number who merely like us. Those early customers give us the best feedback we will ever get, pull the product toward something more people want, and become the foundation for word-of-mouth growth.

Product > sales

The dirty secret of Silicon Valley is that with a good enough sales team, you can probably drag even a crappy product to $1M ARR. That is not the company we want to build. Our standard is simple: we build products we would recommend to our family and friends. The moment we stop doing that, everyone should probably quit. Marketing and sales are critical, but they cannot make customers love something they don’t want. At Replo, product comes first.

Most companies raise too much money

Most fundraising is a waste of time. Most founders fundraise with very little idea of what they plan to do with the money outside of some nebulous intent, like “growing the team.” Investors that get in at a certain price will want to see returns on that price.

Listen to customers, not investors

Investors trade on momentum. Customers trade on reality.

An investor can make a hundred bets and only needs one to work. To them, we're one row in a spreadsheet — a zero is acceptable if another investment returns the fund. You don’t get twenty lives. You’re betting years of your time on one company.

We optimize for customers who really love our product. It’s easier to grow from ten customers who love us to a hundred than to turn a thousand ambivalent customers into evangelists. Fundraising can create momentum for a quarter. Customers who keep paying, complain when the product breaks, and tell their family and friends about Replo create something more durable.

It’s good to be a cockroach

In 2026, we often forget about 2008. Capital markets can disappear overnight, so we constantly ask: At our current revenue growth and spend, will we reach profitability before we run out of money? Fundraising can help us move faster, but our survival should never depend on it. Being a cockroach means staying cheap enough to survive long enough to figure out what the $100B opportunity is. Every dollar of recurring revenue buys us more time, more leverage, and more control over our own destiny.

Why do we think we’ll win?

“Show me the incentive and I’ll show you the outcome.” — Charlie Munger

There are three possible scenarios for Replo’s demise.

The first is that the problem we’re solving doesn’t exist. While I think the probability of this is low for all the reasons above, our response to it is entirely within our control. If we fail to adapt and build something someone wants, that failure is entirely on us.

The second is that another startup comes along and does what we do better than we do. I believe that all startups are built on a secret, and our open secret is that commerce is very poorly understood. Most folks in Silicon Valley think it’s a couple of lines of code integrating with Stripe. When you buy something on Amazon, there is a deep, deep bench of knowledge — from advertising and marketing know-how to credit card routing and logistics — behind each decision.

The third is that a larger incumbent with stronger network effects does what we do. This is the familiar question: “Why doesn’t Google just do this?”

The cost of building software is rapidly approaching zero. In the last generation, companies needed to raise an enormous amount of capital to hire product builders. Mundane, stupid features at scale often took hundreds, if not thousands, of people to build.

To shortcut this, companies like Salesforce, Etsy, and Square built massive ecosystems of partnerships and bundled products together to go to market. On one hand, the advantages of these network effects are obvious. On the other hand, they prevent companies from innovating in areas from which they already benefit.

A large company is not run by a god-king. It is run by hundreds of product managers, sales leaders, and investors with different incentives. A new revenue stream that could transform Replo may be irrelevant — or threatening — to an incumbent. OpenAI, for example, would never launch its own social shopping app if it jeopardized its partnership with TikTok.

We’re willing to go all-in and accept higher risk in exchange for higher reward.

Hiring engineers

What types of engineers are we looking for?

Our goal is to build a cracked, in-person team of engineers in San Francisco. In general, we try to hire people who are better than us.

At this stage, we’re not looking for deep expertise in one particular subject, though it can be helpful in areas like payments, sandbox infrastructure, and inference. We ARE looking for people with first-principles thinking, solid system design, and a passion for the craft of software engineering.

You should be able to work on many problems at the same time and make good trade-offs between product, architecture, and tech debt. Above all, you should take work off other people’s plates and have a multiplier effect on the team.

What types of projects will engineers work on?

We expect engineers to own features end-to-end, including product, design, and customer success. We split projects into programs and platforms. Programs are customer-facing bets that require deep customer understanding. Platforms are shared systems that make many programs possible.

Platform

  • Building sandbox infrastructure for our agents (managing secrets, proxies, and compute)
  • Building an eval framework for our models
  • Launching an analytics platform on ClickHouse/MSK that consumes millions of events per day
  • Building an agentic workflow orchestration architecture
  • Building a public API for Replo/Slack bots
  • Overhauling our billing system to support arbitrary usage-based billing

Programs

  • Enabling customers to buy domains directly through Replo
  • Building agent-driven insight reporting for businesses
  • Building a full product-customizer UI for custom print-on-demand products

What tools do we use, and how does the team work?

We have one-week sprints and a 10-minute, team-wide stand-up every morning. We try to minimize meetings for engineers because every engineer should be on a Maker’s Schedule by default. Everyone is responsible for planning their own roadmap and goals each week.

We aggressively use agents to automate things like first-pass code reviews, Sentry triaging, Slack announcements, first-touch customer support, etc. Everyone at our company writes code with Cursor and Claude Code. Keeping all of our company’s context in code — and making it accessible through MCPs — is key to making us hyper-efficient.

What is our tech stack?

We use TypeScript almost exclusively, with a bit of Python for ML explorations. We use Sentry, Grafana, and PostHog for logging and alerting. We have a proprietary harness inspired by OpenCode, and we aggressively use tools like Braintrust and OpenRouter. Our analytics products use Kafka and ClickHouse.

FAQ

What’s our hiring philosophy?

I’m never in a hurry to hire anyone. A small team of cracked people will beat a large team every single time.

Noah and I have made the mistake of hiring someone because we had a lot of work to do. When we do, we inevitably realize that no one is happy micromanaging them, and things don’t work out.

A lot of people also ask what we’re looking for — this is Noah’s favorite diagram. We aim to exclusively hire people who can identify the right problems, solve them autonomously, and keep the team informed as they do it. This requires sound judgment, strong technical fundamentals, curiosity, and follow-through. You need to know when to move scrappily, when to slow down and be rigorous, and which problems are not worth solving at all.

↑ More money, impact, and freedom

  1. Level 5Superstar

    They identify the right problem and get it solved.

  2. Level 4Systems Thinker

    Tell them the problem, they set up a system to figure it out (people + process).

  3. Level 3Problem Solver

    Tell them what to do, they figure out the how.

  4. Level 2Task Monkey

    Tell them what to do AND how to do it AND when to do it — and it'll get done.

  5. Level 1Useless

    Even if you tell them what to do, they rarely do it right.

What perks and benefits do we offer?

If perks are your priority, don’t work here. We do offer a few practical ones:

  1. As many computers as you need to run your AI agents
  2. $200/month SaaS/book credit, to explore new tools and learn new things
  3. Unlimited AI tokens to do your best work
  4. Free lunch and dinner in the office every day (from real restaurants, not unhealthy catered food)
  5. A free Uber ride home if you are working late

What is our healthcare policy?

We cover your dental, vision, and healthcare, and provide a 401K plan through Human Interest.

What is our vacation policy?

3–4 weeks/year + holidays.

It’s technically unlimited, but we tell everyone to take 3–4 weeks because otherwise no one will take enough time off. To set a good example, Noah and I take 4 weeks off every year (though we also probably work 80–100 hours/week in between).

Still reading and excited? Come find our open roles at replo.app/careers, or shoot me a note at yuxin AT replo DOT app.