Despite being CTO at my startup, I've spent most of my time on non-programming tasks. Every startup looks different, but I'd bet most engineers and researchers don't have a clear idea of what it's actually like to be a founder. It surprised me too, even after reading countless books, listening to countless podcasts, and starting several (less serious) startups over the last 8 years.

Given that many friends have asked about starting startups, I thought I'd cover some things that are either worth repeating or haven't been widely discussed elsewhere.

Quit your job last

Following the bleeding edge of your field shows you both what's possible and the problems people face. I do this by listening to podcasts, reading technical blogs, and following popular people on Twitter. This will look different for you, but what matters is that you're where the experts in your field share what they know.

Along with consuming knowledge, doing projects and then sharing them will help your co-founder hunt and teach you about problems in your field. Many founders I've talked to built a reputation as people who liked to build things, which naturally attracted co-founders. Working on these projects also exposes you to real problems other people face, some big enough to build a company around.

If you're considering co-founding with a friend, run a low-stakes, month-long project with them first. A month is long enough to see how they handle stress, manage time and a project, whether they have agency, and any dealbreakers.

If you have any say in where you live, now or in your next job, try to live in a big city. It widens your pool of co-founders and gives you more chances to run into investors in casual settings. The latter has been a blessing for me, since I didn't come from a target school for investors.

Before you quit, consider whether you can build your startup part-time and without venture capital (also known as bootstrapping). Until my most recent startup, every one I started was something I did while in school or working. Bootstrapping makes it far more likely you'll build a "4-Hour Workweek" kind of company. Some ideas, though, need speed and capital to be possible, and for those you'll need venture capital.

If you do want to raise venture capital, keep a list of accelerator deadlines so you can apply, and ideally quit your job once one accepts you. Many firms also publish Requests for Startups: lists of ideas they want funded, which can be a useful place to start. A few well-known accelerators: YC (Y Combinator), Antler, SPC (South Park Commons), and Neo.

The idea is the easy part

Finding your first paying customer is the real test of whether an idea is workable. I had plenty of interesting ideas, but for many of them, my lack of credentials or connections would've made getting customers near impossible.

This means you have less freedom in what you can work on than you'd think. Some ideas are just better pursued from inside an established company, and if you haven't worked at the best place for that kind of thing, it might be worth doing that first.

As soon as you've anchored on an idea, identify a few ideal customer profiles (ICPs) and try to schedule discovery calls with them as soon as possible, even before building the product. The Mom Test, a short book on how to interview customers without fooling yourself, lays out a great way to run these calls.

Selling is your new job

Whatever your role, every early-stage founder has to learn to sell, whether it's the company, themselves, or the product. There's no shortcut other than doing it, though books help.

Your website is your most important passive salesperson. Make sure it looks good and is optimized for the right profile: investors, customers, or new hires.

When you're selling actively to customers, first figure out whether they're small-to-medium businesses (SMBs) or enterprise, because it changes everything downstream. SMB tends to be faster, lower-touch, and close to fixed pricing; enterprise is slower, relationship-heavy, and negotiated deal by deal.

Fundraising should be a sprint

Aim to finish fundraising in as short a window as possible (~2 weeks), because it eats up all of a founder's time. To pull that off and get your best terms, you want to control your narrative:

First, don't talk to investors until the sprint starts. You're unlikely to have your pitch materials down before then, and talking early doesn't seem to improve your odds.

Second, don't share your pitch deck, since it tends to circulate. There are exceptions where you'd talk to an investor early, say one who's well aligned with your vision, but even then, keep the deck to yourself.

Third, staying in stealth keeps competitors and investors from forming an opinion before you're ready to set one.

Once the sprint starts and you're actually pitching, make sure you have a coherent story. Put simply: investors want a market that's big enough, a product that's magical, differentiated, and defensible, the right team, and a clear way to make money.

Finally, one note on valuation that people overlook: raising at a high valuation narrows your acquisition options down the line, since any acquirer has to clear the multiple your investors need.

Most of the job isn't building

Even as the CTO, most of my time isn't spent building.

A lot of it is writing. Internal strategy, proposals to win contracts, that sort of thing. The good part is it doesn't go to waste: those docs have become our knowledge base and a record of why we made the calls we did, and I can feed them straight to our LLMs.

You'll also end up having to form opinions on things you know nothing about, then act on them. I designed our sales playbook and decided what our sales materials should look like, none of which I'd done before.

And through all of it, keep the long-term vision in front of you while assuming most of what you try won't work out. As Andy Grove put it, only the paranoid survive.

Appendix: random tips that I've found useful

• Read competitors' job descriptions to see where they're headed and what they're building.
• Read your customers' job descriptions to spot when they're hiring for something your product could do instead.
• Push off hiring until you actually need to. When you do, hire superstars, and hire slow, fire fast.
• Advisors: message people, chat a few times, and if it clicks, ask them. Expect to give 0.1-0.25% equity or pay cash. Know the difference between advisors you bring on for clout and ones who actually understand your business.
• Advisors and customers will sometimes ghost you. It almost always just means they're busy, not that you did something wrong.