Comment by mooreds | Hacker News Reader

Comment by mooreds | Hacker News Reader

Parent Full thread

Disagree.

I helped found a startup[0] and have been around a number of them, in slacks and on email lists[1].

I think that domain knowledge is so so important for startup success. Even more important than technical expertise.

Why? Because the main problems for 0->1 startups are business problems, not technical ones:

* where do possible customers hang out

* what problems will they pay you to solve

So in that case, a good CEO who knows the domain is critical. They'll have good answers to these questions, which will help guide the technical decisions. Or they'll know who to talk to in order to find the answers. There's no shortcut to that knowledge; you have to spend years to get experience in the domain. Learning and reading about the domain can help, but you still have to spend the time.

The main exception, and it's a big one, is if you are building something in the devtools space, where you can be your own customer.

0: https://www.thefoodcorridor.com/ (I left after a few years, but they are still going strong!)

1: https://ctolunches.com/ (free email list for engineering leaders, so many good discussions happen there)

100% agree. I'm a CTO for a vertical SaaS, working with a first-time founder with deep domain expertise. He came in with a thesis as to why our target customer was still using Excel, even though it was completely unsuited to their needs. Almost everything he thought at the beginning has been borne out, and there's no way I would have figured out this market on my own. There's also no way he would have built a usable product without someone like me - his original plan was bloated and we narrowed it to the core where we have very specific value.

I've been a solo founder as well multiple time co-founder / CTO at venture-backed startups so I'm certainly capable of starting something myself, but I think this combination is an excellent model. A couple of other things that make it work: the CEO has sales skills, and does not have much of an ego and is willing to listen to someone who has been through all of this multiple times.

> A couple of other things that make it work: the CEO has sales skills, and does not have much of an ego and is willing to listen to someone who has been through all of this multiple times.

Yeah! You definitely need to have a good CEO as a partner. A bad CEO is worse than no partner at all.

Those are some great attributes of what makes a good CEO and partner.

I have hardly come across aspiring-CEOs who do not have much ego and listen to other founders, let alone founders or serial-founders. After some founder dating last year, I am a bit jaded, though.

yup

Also speaking from similiar experience, ^^^ this is the best combo. Where both parties have a certain degree of humility and self-awareness and understand how their skills, personalities, and personal circumstances (ability to travel, life stage, etc.) compliment one another.

Right, sure, there is an initial pain of building domain knowledge. But that one year of knowledge building, distribution, early customers, then dominates who has control over the cap table for a decade. Why not just do this work yourself? Its some weird myth that someone capable of being a technical leader cannot also flesh out the distribution

Especially with the existence of AI, it really makes no sense and is some weird hierarchical system built by business people

> But that one year of knowledge building, distribution, early customers, then dominates who has control over the cap table for a decade.

Can you recommend any resources for learning how to do this work yourself?

One shortcut would be to:

1. pick a vertical (construction, legal, accounting, etc)

2. look at how to apply ai (compliance, voice customer service, email)

3. find sales people from a would be competitor or old school b2b saas in that area, and pay them 150$ an hour to tell you what customers are paying for

4. at the direction of how to do sales, come up with the product and either do the sales yourself without building, or pay someone to do cold calls and pitch the product

5. build the product

I'd recommend:

- the lean startup: https://www.amazon.com/Lean-Startup-Entrepreneurs-Continuous...

- Amy Hoy's blog: https://stackingthebricks.com/

- working in the domain you are interested in for a year. Go be a realtor. Go work in a lawncare business. etc. etc.

- read Patrick's advice about how he found customers for Appointment reminder: https://www.kalzumeus.com/2010/05/14/unveiling-my-second-pro...

edit: added link to Patrick's piece

Maybe AI has changed things, but I think it takes more than one year to gain sufficient domain expertise.

But your argument still has validity even if it takes 2-3 years to gain the knowledge, so let me address that.

My response is: if I wanted to be an expert in <startup domain> to the point I didn't need a co-founder CEO, I would take the time, the years I mention above. But that has costs (opportunity costs if nothing else).

Instead, I prefer to be an expert in software development. This gives me less control in each given startup as you point out, but more optionality across startups (and other companies)[0]. This choice also means that I give input on the hard decisions that company leadership has to make, but I don't have to make the final call. Frankly, that's more weight than I care to bear.

That's a personal choice. I've watched friends be CEOs of successful startups and it's not always (nor usually, even) fun or easy.

0: I wrote more about that optionality here: https://www.mooreds.com/wordpress/archives/3445

You're trying to have a serious discussion with someone convinced he can build deep domain knowledge in eg banks and their needs, how to sell to them, plus the associated network, in a year while also writing code. Oh, and just assumes "early customers", ie you're also going to learn how to sell. Because how hard can it be :rolleyes:

> Its some weird myth that someone capable of being a technical leader cannot also flesh out the distribution

This idea often gets repeated here but it’s based on ZIRP from 2010s driving up developer salaries to a point where every kid with some portion of a CS degree thought he was the next Zuckerberg. “Technical” contributors became convinced they were the only contributors. Lots of stories about “idea guys” and useless upper management who Just Don’t Get It (meanwhile most guys posting things like this have never even seen a balance sheet). It’s a popular idea here because it establishes devs as the smartest guys in the room, but it doesn’t have a lot of bearing on reality. Plenty of devs have neither the aptitude nor inclination to learn the business side of things, as evidenced by some of the lifers who post here about upper management.

This attitude will change as market cools off, no code and LLM output improves, and overseas contractors become more accessible.

> Especially with the existence of AI, it really makes no sense and is some weird hierarchical system built by business people

LLM code output can be tested as it is generated, whereas its advice on accounting and contract law can only be tested by domain experts, who you can just pay for this advice without involving the LLM.

ycombinator, the best very early stage investors of all time hold this opinion:

"If you are a technical founder, you do not need a non-technical cofounder."

https://x.com/snowmaker/status/1948160642399314188

Its really not up for debate, at this point it is a borderline scam to work for a CEO that isnt significantly better/bringing in capital. You are basically giving them free labor

You don't need a non-technical cofounder, but you do need a founder willing and capable of handling non-technical things.

If nobody in your startup team wants to handle sales, marketing, finances, and business operations, you're going to end up with working software that nobody's paying for.

Jared has an (unfinished?) degree in CS and Econ so that’s a bit rich for him to say. Why’d he bother with Econ if it’s so UsELeSs?

What’s YC’s track record in the new LLMs-can-code era?

Or are we talking about Dropbox and Airbnb?

most of the recent series A ai companies are YC companies, they are absolutely crushing it

What are their names?

>at this point it is a borderline scam to work for a CEO that isnt significantly better/bringing in capital

I never contested that. My point is that there are plenty of devs who think they should found a startup based on the mistaken idea that they can learn the business side of things overnight.

If you find mentors, or just pay people 200 an hour to meet once a week, a lot could be learned. I agree the gap is large, but it can be overcome

>LLM code output can be tested as it is generated, whereas its advice on accounting and contract law can only be tested by domain experts, who you can just pay for this advice without involving the LLM.

Assuming that passing some tests means it’s good code is as dumb as asking the LLM accounting and contract law advice.

It can be superficially correct and completely broken.

> It can be superficially correct and completely broken.

This also applies to developer output. If you don’t have someone qualified in the loop to foresee problems and perform testing, you could end up with something that falls over later on, but by the time that is consequential you can pay someone to do that kind of testing. By contrast, a dev who decides to be his own lawyer only gets to test the validity of his contracts if he hires a lawyer or gets sued.

That same thing happens with accountants and lawyers. I’ve dealt with terrible of both that were just good enough to not get disbarred or lose their license.

I don’t think you have a salient point here.

> I’ve dealt with terrible of both that were just good enough to not get disbarred or lose their license.

The existence of bad surgeons doesn’t make performing surgery on yourself the better alternative.

A sufficiently-motivated analyst can hack together an app and test it as he goes. Yes, there might be scaling issues and edge cases, but he’s putting together a proof of concept; once he has validated the market, the developers who build the app “correctly” are effectively interchangeable. There are questions of professional ethics if the software is touching anything to do with medicine, personal information, or other sensitive issues like that, but that does not represent a majority of development. If you mean “good” in the sense of being artful or efficient, then no one cares.

> This idea often gets repeated here but it’s based on ZIRP from 2010s ...

I don't think it is just the ZIRP/2010s. Here's a post I wrote about developer arrogance in 2004, well before the ZIRP era: https://www.mooreds.com/wordpress/archives/134

The best combination is a technical CEO founder with deep domain expertise. Confirms what you say about the dev tools space exception.

Of which LLMs are effectively a dev tool that happens to also be a tool for every other industry vertical in the world economy.

Why can't everything you just described be done by someone who doesn't have the CEO legal title? The CTO can be the CEO in title, but fulfill the duties of both, with consultation from the "CEO" (pick whatever title you want). Basically inverting the non-technical-founder / CTO relationship, in favor of the technical founder who needs non-technical expertise and guidance, but still wants to maintain overall control of the business.

Just remember that every startup is not a SaaS business...

Reply on news.ycombinator.com