Things I didn’t understand about starting a company until I tried

December 12, 2020

At the end of August, my friend Pat and I decided to work on an idea he’d had: to create a company providing work from home employees with the perks that they’d enjoyed at their offices. At the start of October we launched the platform, Cholla. You can read more about the start of our journey here and here

We did lots of stuff right. We’re both avid consumers of startup lore – Paul Graham essays, podcasts, and books. We knew all of The Hard Thing about Hard Things,  and we’d watched lots of the content from YC’s Startup School. I’d give us a solid 8/10 for avoiding most of the common pitfalls that first time founders fall into. This post is about some of the stuff that I still didn’t understand until I’d actually tried it.

Some background

Pat and I met working at a highly capitalized AI startup called Element AI, where we learned how to build software products. EAI had many good features and some bad ones, which are currently being pored over by many armchair critics in the wake of the company’s sale to ServiceNow.

One of the external dynamics with which Element struggled was the peculiar incentive structure currently afflicting applied machine learning. AI talent, as everybody knows, is scarce and highly paid. What’s less discussed, however, is that the incentive structure nudges people towards fundamental research, as opposed to applied work that directly impacts customers. A reliable way for a math-savvy software developer to double their salary is to migrate first into applied research – producing models which can be used in a business context – and then into pure research, where their focus is on publishing papers.

In fact, it’s even worse than that. Papers, unlike software, are permanently prestigious and open source. The hundreds of thousands of lines of code we wrote for Element AI in the pursuit of making products won’t ever be visible to the rest of the world (particularly if our product gets canned), whilst every paper published remains in the public domain. So not only can you get paid more for doing fundamental research, you get to keep your work in perpetuity.

This might seem odd, and it is – my pet theory is that this is Google and Facebook’s brilliant way of preventing small companies building valuable AI products, because everybody is more richly rewarded for publishing papers. But that’s a different blogpost. The relevance to this story is that both Pat and I emerged from Element AI with an aversion to seeking out ‘interesting and publishable’ work. We focused on solving the frequently messy and boring enterprise problems that we regarded as the bread and butter of our company’s existence, and we regarded this as a more honourable – if ultimately less remunerative – practice.

So when Pat came to me with the idea of starting a company shipping people coffee beans and cereal bars, my first reaction wasn’t ‘hmm, that doesn’t sound like something that somebody with a PhD in neuroscience should be doing’, it was ‘yeah I see how we could make that work.’ We both, to varying extents, self-identified as pragmatic grifters, and it was clear that grift and some standard software engineering could get this thing off the ground.

Product founder market fit

Many of my friends, however, did think this was a bit odd. I got bemused smiles when I told people I was spending my time compiling lists of muesli companies. People were slightly surprised that I was worrying about Zoom yoga rather than working on some pressing societal issue. Shellacked in startup folklore, I didn’t think too much of it, dismissing it as the standard skepticism of the non-entrepreneur. Nobody we spoke to actually had a good explanation of why this was a bad thing to build, so we forged on.

As it turned out, this wasn’t a bad thing to build: it was a bad thing for me to build, and that’s how I should have interpreted some of the skepticism. This was a surprising thing to do given my previous interests, and the way that this played out wasn’t in the viability of the company, but in the durability of my enthusiasm. I had spent so long judging people who sought out ‘interesting’ work that I dramatically  underestimated  how important it was to me.

Pat and I were both well versed in Marc Andreesen’s characterization of ‘product market fit’ as the ultimate goal for any early stage company. We worked hard on this, and avoided building something nobody wanted. But I missed the extension of this idea, emphasized by Naval Ravikant: product-market-founder fit. If the founder doesn’t have the skills and the passion to execute on an idea, it’s not going to work out.  The best companies are a distillation of their founder’s personality- like Steve Job’s perfectionism at Apple and Musk’s cocky ubergeekery at SpaceX.

We has discussed and dismissed our lack of domain expertise as a barrier. And we were correct: lack of expertise in the space wasn’t a critical impediment to building the platform. But being a domain expert doesn’t just articulate your knowledge of a domain; it’s proof that the domain interests you enough to become expert in it. If you’ve been working on a topic for 10 years, there’s a good chance you’ll still want to work on it in 3 years time. But if you’ve only been thinking about something for a month, the odds that you’ll still be thinking about it in a year aren’t great.

Investors offer more than money

Another dramatic recalibration that Pat and I underwent on our journey was with respect to venture capital. Once again, our opinions had been heavily influenced by Element AI, where a surfeit of capital facilitated explosive growth and a chronic lack of focus. ‘No VC for us’, we told each other. ‘We’ll keep our incentives aligned with our customers by bootstrapping all the way’. Within a couple of weeks of starting, we both had far more nuanced opinions.

It turns out that there are lots of reasons you might want to share your business with a group of external investors, and all of these are way more pertinent to you when you’re actually founding companies than working in them. On paper these are glaringly obvious, and please feel free to roll your eyes at my naïveté.

The most intuitive reason to take funding is financial: you get to reduce your personal risk by having somebody else finance the business. In return, you sacrifice some of your upside. This is a fantastic mechanism because otherwise aspiring entrepreneurs would frequently get wiped out by their first failed venture and wouldn’t go on to build valuable companies. Think of this like the soda to your scotch. You might prefer neat scotch, but you’re going to be able to have many more drinks without ending up on the floor if you dilute it.

We weren’t starting a rocket company so I had thought we wouldn’t need much cash. But your time is also an investment, and taking capital spreads that risk too- because you get to pay yourself. Both Pat and I were earning pretty good money before embarking on this ride, so the effective cost of building this thing was several thousands of dollars a month. That’s risk you can share if you take external financing.

Secondly,  investors have seen a lot of startups, and they are better placed than anybody to evaluate your idea. That’s literally their job. By opting out of VC, you’re opting out a rich pool of advice, expertise, and networking that a good VC can provide. Unlike many of the people you talk to about your idea,  they’re actually likely to tell you the truth, because it’s easy to tell if they think your idea is crap- they won’t give you money.

The third reason sounds sort of pathetic, but I think it matters. When you’re very early stage, the list of people who really care about the fate of the company is you and your cofounders. When VCs invest, they’re adding their names to that list. Sometimes it’s just really nice to have some older, experienced people on your team (who you aren’t paying with your limited capital). The morale-boost associated with VC (“look, somebody rich thinks this isn’t insane!”) does a little to thicken the skin of your fragile new company.


Pat and I had a tremendous time getting Cholla off the ground, hacking together infrastructure and talking to lots of early customers. After 6 weeks or so I decided to bow out and focus on climate change work, which increasingly felt like the subject I wanted to devote the next several years of my career to.

Happily, Pat was joined by some other ex Element AI employees whose aptitudes and appetites were far better suited to running an events company than mine. These included our fantastic events specialist, Guillaume Gagnon. They’ve run more than 40 events, for over 600 people, generating tens of thousands of dollars in revenue.

When I left Element AI, a great mentor told me that “figuring out what you want” mattered far more than “getting what you want”. In that sense, Cholla was a resounding success. I look forward to future chapters in entrepreneurship, and I hope that I’ll be lucky enough to find anyone as formidable as Pat with whom to take the plunge.

© 2023, Built with Gatsby