21 things I learned (so far) about building a company

Ever since starting college, I've been drawn to the idea of building a startup. I joined a few founder and VC communities here and there to explore this interest, but they were mostly either surface-level introductions, or programs I didn't fully commit to in favor of other activities.

I wanted to experience what building a company was like, so when I heard that YC was launching a program at MIT this January, I got together with two of my friends and applied. The premise of the program is straightforward: spend the first two weeks building an MVP, then the next two weeks getting a paying customer. Once a week, the YC partners would hold a group session to speak about their experiences and lead discussions around each team's challenges.

I found a lot of their takes quite insightful, so I wanted to compile my takeaways here for future reference!

Founding and Teams

1. Assume that your first idea is wrong.

Your idea will change. This means that the real commitment you're making is to your team. Ask yourself, "would I enjoy spending 18 hours a day with this person?" and "would I still want to build something together even if this idea fails?"

To minimize risk, choose people you've actually worked with before, and whose work habits, stress reactions, and conflict styles you already know.

2. YC is a one-way door.

Your startup will likely fail if you're not fully committed, both physically (no side job, all working hours go to the startup) and emotionally (no backup offer waiting in your back pocket). Go in with the mindset that you're not returning to school, your job, or your internship. If you're keeping one foot out the door, you're not ready to walk through it.

3. Decide roles early on.

In the early stages, you and your cofounders will work on everything together. However, it's important to decide upfront who will call the shots when there's a disagreement.

4. Hiring and keeping good people is one of the hardest problems.

Finding good people is hard, especially since they will often have better offers. One approach for filtering for people who are genuinely interested is to create a small challenge or project that tests for the skills you actually need.

Building and Product

5. Don't build in stealth.

Launch the worst version you can get away with. Showing it to customers and getting them to pay will give you signal that building in private won't.

6. Don't be afraid to overfit to one customer.

When landing a first customer, it can feel like you're tailoring too much to their needs (e.g. building niche features, doing manual tasks for them) rather than solving a more general problem. This is okay. As Paul Graham said, Do things that don't scale first. Keep talking to more customers to collect more data points on what's actually niche vs what's universal. What might feel like overfitting could actually be a common pattern.

7. Get a dollar value on each feature.

Customers can often ask for features that they don't actually need just to have something to say. To determine which requests are worth prioritizing, ask "How exactly would this feature help you? For example, would it increase revenue by X? Reduce costs by Y?" If they can't answer, it's probably a nice-to-have, not a must-have.

8. Engineering taste is your moat.

Why can't companies just build their own internal tools with Claude Code? Even if big companies could coordinate well enough to do it, and small companies had the time, your edge is taste - knowing what to build, making it intuitive, making it feel good to use. If your product is something a customer could easily prompt an AI to make themselves, you need to offer something more: better design, deeper integrations, or domain expertise they don't have.

9. Your startup needs to fall into one of two pricing camps.

If you're charging <$10K/year per customer, you need product-led growth - a product so intuitive people adopt it without a salesperson. You'll need 10,000+ customers to reach $100M ARR (think Slack, Cursor).

If you're charging >$100K/year per customer, you can have a sales team selling directly. You only need ~1,000 customers to hit $100M ARR (think Salesforce, Oracle). Almost all money in B2B startups is made in this category.

Customers and Sales

10. Customers who use your product for free are not your target customer.

They're fundamentally different from those who pay. Free users don't give you real signal, and they may simply say nice things because it costs them nothing.

11. Only 10% of potential customers are early adopters.

If you pitch 10 people and they all pass, that doesn't mean your idea is invalidated - you probably just haven't found your early adopter yet. Don't doubt your product until you've talked to ~20 potential customers and gotten rejected by all of them.

12. Find your "champion" - the person who gets promoted vs loses sleep over your product.

Sales close because someone inside the company needs your product to succeed, not necessarily because your product is technically impressive. Find the person whose career depends on solving the problem you solve and let them fight for you internally.

13. In big company sales, technical quality of your product matters less than you'd think.

Instead, they depend much more on: (a) timing, (b) knowing the right person, and (c) building something that makes the difference between your champion (see previous point) getting promoted or losing sleep.

14. Watch how your customer reacts during demos.

Good signs: they ask to use the product themselves, they start thinking out loud about how it would fit into their workflow, they try to negotiate the price down. Bad signs: they nod politely, say it's cool, and don't ask any follow-up questions. Learn to tell the difference between genuine excitement and just being nice.

15. Be opinionated about how you work with customers.

When talking to customers, be firm about your price and pricing model. You can even give them a structured plan - e.g., weeks 1-3 we'll implement X, month 2 we'll do Y, month 3 we'll do Z. Each phase has a set price, and they can opt out anytime - but if they leave after week 3, they still pay for weeks 1-3 because you put in the work.

16. Offer a contract where they pay up front but can get a refund.

Getting early customers to pay can be difficult. One approach is to guarantee a refund if they're not happy after 30 days. This way, they still go through the process of actually paying - getting budget approval, signing the contract, and mentally committing to the purchase.

17. If they agree to your price immediately, you undercharged.

The sweet spot is when they start negotiating. One YC partner's advice was that "every time you undercharge, double the price next time until you get pushback." Discover the optimal pricing for your product through experimentation.

18. Create a false sense of scarcity.

You can say something like "we can only take on 2-3 companies in the next few weeks, so let us know before then if you'd be interested." This forces the customer to make a decision.

Competition

19. 50% of your competitors will die from cofounder conflicts.

"Don't die" is a real strategy. If you can stay afloat long enough, it's likely that your competition will fall apart from internal conflict. In the same vein, you're far more likely to die from self-inflicted wounds than from competitors.

20. Competition rarely kills startups.

For example, when Charles Schwab (one of the largest brokerage firms in the U.S.) launched their own robo-advisor - an automated investment tool - smaller robo-advisor startups assumed they were dead. However, Schwab entering the space actually validated the market. Other financial platforms saw this and started reaching out to those startups to partner.

21. If the pie is big enough, competition doesn't matter.

Most startups die because they didn't make the pie bigger, not because they lost to competitors. Apple didn't win mobile phones by taking Nokia's customers - they expanded the entire market.

Being a part of this program has excited my inner child more than anything has in a while. As someone who gets bored easily, this program offered exactly what I needed - an opportunity to learn, build, and fail quickly. I definitely plan to take these learnings with me wherever I go next.