We got into YC. We turned it down.
My story realizing and renouncing a childhood dream
Welcome
I’m Anirudh.
The entirety of Miracle (YC W23) runs off of my code (as of May 2024). It’s used by hundreds of clinical professionals and pharma executives daily to extract key real-time insights into the progress of their clinical trials.
Previously, I was an early engineer and the second ML hire at Whatnot (YC W20, ~$4B unicorn) primarily focused on content discovery and recommender systems.
Ever since high school I’ve wanted to be a startup founder, and twice a year, I would scour the YC launches and acceptance posts and pray that one day my name and company would be up there with them. This is the story of how that dream came true in the most unexpected way.
What is YC?
Quick context: Y Combinator (YC) is a legendary startup accelerator started by Jessica Livingston and Paul Graham. OpenAI CEO Sam Altman was previously its head. It has been the first check in startups like Airbnb, Dropbox, Stripe, Reddit, Instacart, DoorDash, and Coinbase. YC startups have a combined $600 billion valuation.
The standard deal is $500,000 for at least 7% of your company, and most YC companies raise additional funding during the biannual, three-month program. Participants typically receive $1-3m each in total seed funding from world-class investors in addition to boatloads of community support including YC-specific software discounts, an internal YC social network called Bookface, and a YC jobs board.
My Application
My original idea was called Poof, which was a way to turn your Instagram bookmarks into a Pinterest-like shoppable experience.
Having worked at Whatnot and StockX, I saw firsthand how hard the cold-start problem was to overcome, and how it’s generally really hard for consumers to find stuff they like. The thing is, we have data on what consumers like (IG/Pinterest), and we have ways to buy those things (Depop, Poshmark, etc.) — the challenge is connecting the two dots as seamlessly as possible, which is what I attempted to do with Poof.
What is your company going to make? Please describe your product and what it does or will do (YC Application)
Poof allows a user to DM posts to the Poof Instagram account the way a user would share posts with their friends. In addition, the user can share an entire bookmarks folder with Poof without leaving the Instagram app. Extending this experience to different apps allows Poof to innately capture valuable taste data without detracting the user from their normal activities or wasting precious time in onboarding with auth flows or capturing extraneous data. Compiling these data points, Poof scrapes eBay, Poshmark, Depop, etc. to find sustainable clothing options that match the user's taste while practically eliminating the traditional consumer onboarding process.
On April 26th (my birthday), I got this email from Diana Hu, one of the YC group partners. This felt surreal already, just kind of acknowledging that the YC group partners are real people and can talk to us common folk.
Basically, however, she was telling me to get a cofounder if I wanted to have a real shot at YC. While YC accepts solo founders, their strong advice is that teams are better than individuals.
Joining Forces
I knew my friend Svapnil Ankolkar had quit his job at Modern Treasury (YC S18) a few months ago, and he had already built Presidential AI, a Discord bot that scaled to 50,000 users and profitability. We had both worked for YC companies, we had a lot of mutual friends in the startup/tech space, and we agreed on the fact that you can build full-featured apps inside of existing consumer infra.
We agreed on a “hack the group chat” wedge to grow a consumer app, albeit mine was via Instagram, and his was via Discord. We did (and still do) strongly believe that this approach allows you to build faster, iterate faster, and grow faster than having to build and publish standalone apps and spend valuable resources into user acquisition and retention.
I really liked that he, too, was zigging when everyone else was zagging. He was building something in crypto (not AI) and building a Discord bot. Nobody was doing both of those things. However, I knew nothing about crypto nor Discord, so I was naturally hesitant to commit to a venture-scale business on this premise.
Nevertheless, neither of us knew anyone else who was building something similar enough, and we figured adding each other as cofounders could only help with the YC application.
We added each other to our respective applications, and we crossed our fingers.
The YC Interview
We kept building independently and keeping in loose contact, and I hadn’t quit my job yet. Two weeks later, on May 12 (a Sunday), we get an email saying we’d been scheduled for a YC interview on May 14 (a Tuesday). Naturally, this caught us pretty off guard, since we’d been seeing YC interviews and acceptances all of the past week on Twitter, and now we had less than 48 hours to prepare and give our best effort.
During this time, Poof had been deplatformed by Facebook, meaning I couldn’t make the bot publicly available. This was clearly a significant hindrance. I thought about making it a Discord or Telegram bot, but my target audience (college students/new grads, primarily female) were not known for being active on either platform.
This and the time pressure meant we had to roll with Svapnil’s application, which was Moonbot, the fastest way to trade memecoins on Discord. We spent the entire time leading up to the interview doing the iPG Simulator to practice commonly-asked YC questions and our answers with a timer.
We did practice rounds with our founder friends, as well as practice rounds with people completely removed from the tech/startup space, just to practice our communication skills to non-technical people.
The YC interview is only ten minutes long. You have to give your answers as concisely as possible while sounding natural and non-scripted. Doing as much practice as possible can only help. All you need to do is clearly explain what you’re doing, who you’re building it for, and how you’re generating revenue. As corollaries to these points, you should have tangible numbers and user stories. You should talk to users and be able to give direct quotes from users. You should have dashboards that keep track of key metrics and know your key metrics off the tip of your tongue. Avoid using jargon unless absolutely necessary.
Practice explaining your idea to non-technical, non-startup friends. If they can get excited, you should be golden.
Caveat here: avoid tarpits, which are apps that are seemingly obviously great ideas that have been tried and failed a number of times.
After the Interview
The ten minutes go by, and we feel like we actually do surprisingly well. We knew our metrics, defended technical decisions, and defended non-technical decisions by bringing up real user stories.
Around 10pm PST the same day, our group partner, Brad Flora (first check into OpenSea, Whatnot, and RazorPay) sends us a Zoom link. We’re incredibly nervous, but hopeful. He tells us that we’re officially in YC, but not before he asks us the real and tough question, who’s in charge?
We had prepared this answer for our interview, saying Svap would be in charge of product and strategy, while I would lead engineering. However, the reality was I didn’t know anything about Discord or crypto, so I’d be a pretty bad CTO having to make a multitude of technical decisions I was unconfident about. More than anything, Svap and I had never actually worked together. We’d never built anything together, and we were never more than just acquaintances who talked about startups and ideas from time to time.
We ask for a couple days to consider the offer; Brad says “how about a day?” We say okay. The next 24 hours were the most intense 24 hours of my life. I can vividly recount them, hour-to-hour.
I put in my two weeks that day, deciding that I would start my own thing full-time, independent of YC. I spend the entire day simultaneously juggling offboarding and ping-ponging the YC decision by the minute in my head. Offboarding as a solo engineer is a pretty large undertaking in and of itself — grateful my head didn’t combust from the additional pressure of YC.
Turning down YC
At the end of the day, the cold hard truth was that I had not contributed to Moonbot (Svapnil’s project) since I had been working both full-time and on my own idea, and I had a lot of catch-up to play with both Discord and crypto.
While Svapnil and I agreed on the core premise of building rich and immersive chatbots, we were unable to see eye-to-eye on the future of our joint company. The fact that we had never worked together made this even riskier.
Starting a venture-backed startup is serious business. Millions of dollars in funding adds overwhelming pressure to execute and succeed. Founding necessitates a team that will relentlessly pursue a shared vision, independent of any investors.
This whole time, we basically only applied to YC thinking getting in was the hard part (spoiler alert: building a business is harder). We hadn’t really given much consideration to what happens when we actually do get accepted — and the truth is if your co-founder is only building with you because of YC, you’re just not going to make it.
Talking to our founder friends independently, it became more and more clear that “shotgun wedding” YC is not the best idea. Another friend said founding is “like dying over and over again and finding a way to get back up.” If your cofounder isn’t someone you trust with your whole heart in these situations, your startup isn’t going to make it.
Svapnil and I have communicated to each other and our friends over and over again that we believe in each other as both founders and engineers, but that doesn’t inherently mean we would make good co-founders.
Lessons for Readers
These are all mentioned above, but condensed here for convenience:
YC (or any external investment) is the starting point, not the destination. You still have to build a business.
Founding a business is hard. Cofounders need not be best friends, but they must be willing to put aside anything besides family and personal health to pursue identically common goals
Two good founders and engineers may not make the best cofounding pair. Just because you can build something together doesn’t mean you should. You should stress test your chemistry to the extent possible before taking on significant investment.
What’s next?
I’m old enough to have experience, and young enough to be young. I am going to build.
As I mentioned, I am in the process of offboarding at Miracle. It’s nice to not have the pressure of YC or other investment at this time; it allows me to truly build something I care about with people that will work with me for the right reasons.
I’m trying to talk to all my friends and former coworkers and evaluate where their heads are at. I’m hacking on some side projects, seeing what sticks, and generally figuring out what problems people have and how I can potentially fix them.
Between my time at Miracle, Whatnot, Apple, StockX, and more, I’ve worked across a software spectrum spanning consumer, e-commerce, b2b saas, and healthcare. My industry strengths are in consumer and e-commerce, and my technical strengths are in full-stack, data, and ML engineering.
Get in touch
☕️ Let’s get coffee at my favorite bay area spots
If you enjoy tinkering and building and are in the SF Bay Area or greater NYC area, tweet at me @kamathematic.
If you have any ideas for what I could do next, also tweet at me @kamathematic.