Two observations:

1) Shipping-and-iterating is uncomfortably hard. Putting your product out there in front of users is painful. It's a lot easier to just constantly brainstorm ideas or hide in the coding cave.

2) As a founder, it's tempting to think that your startup is unique: what you're building is uniquely challenging, important and highly non-trivial.

Anecdotally, I've seen far too many founders (including my own past self) use 2 as an excuse not to do 1: most of the time, those never even launch. And not just in the hardware space, but also in: health-tech, bio, fin-tech, gov-tech, legal-tech, insur-tech, prop-tech, food-tech, logistics, and even B2B SaaS ("what we're building is hard and needs enterprise-grade robustness / security").

There may be ways to build massively successful companies that don't involve rapid shipping-and-iterating cycles, and in general there may be ways to build successful companies while ignoring or even doing the exact opposite of what YC advises.

But YC would know a few things about hard-tech from having funded possibly hundreds of such companies, including several massively successful companies. In fact, the top 3 YC companies of all-time as of 2020 are all in highly-regulated spaces: Stripe, Airbnb and Cruise [1]. Cruise in particular is a hardware company (self-driving cars) whose billion-dollar success was largely due to their ability to ship-and-iterate much, much faster than pretty much every other self-driving car company out there.

[1] https://www.ycombinator.com/topcompanies

I think being in a highly regulated market is orthogonal to the original point. Regulatory compliance is a checkbox and not a differentiating capability, even if it is an arduous checkbox. Hard/Deep tech is a completely different animal. None of the three YC companies you list fall under this rubric as they are essentially novel presentation layers on existing tech (yes, this includes Cruise, whose tech I am familiar with). The point isn’t about startups where the hardest thing you need to do is business execution and compliance.

Some kinds of hardcore tech, even deep software tech that has no compliance issues, has extremely steep engineering hurdles that must be cleared before you have the most minimal of MVPs that will have any currency with customers. How do you iterate fast when the most minimal demonstration of capability requires writing hundreds of thousands of lines of highly sophisticated code from scratch? Or requires physical engineering that necessarily takes years to complete? Many of these kinds of opportunities exist but almost all of the “ship-and-iterate” guidance is for companies that are using core tech, not creating core tech. That’s playing the startup game on easy mode.

If we end up in a place where a “tech startup” is defined as companies that are technically trivially to execute and amenable to rapid ship-and-iterate, which admittedly is most of them, then we might as well elide “tech” from the nomenclature. At that point, they are merely tech-enabled ordinary businesses, and what ordinary business isn’t tech-enabled these days? If deep tech problems could be solved by slapping together some open source and layering on clever marketing, they wouldn’t be problems.

It feels like we’ve redefined “tech startup” to exclude doing hard technical things, but aren’t honest about that change in definition.

Yes that is great advice for many startups most of the time, but my whole point in the grandparent was that when as a founder you find the hand-me-down wisdom doesn't fit, don't be afraid to leave it alone and think for yourself.

PS. If your criticism were right my coding cave is a 1000m2 factory, which has to be some kind of record. /s

In your parent comment you mention:

> not much of the strategy content applies to hardware companies

While it's probably a good idea to not blindly follow every single piece of advice (by YC or anyone), saying that not much of the content applies is probably a bit of a stretch. Considering how many successful companies YC has funded in the hard-tech space, I'd be receptive to at least some of their advice / insights.

You also mention:

> you have this de-facto "ship and iterate" model from the SaaS world which just doesn't apply to nontrivial hardware projects.

Again, it may be a good idea to question the advice of "ship and iterate", but saying _it just doesn't apply to nontrivial hardware project_ is again a bit of a stretch. As the above commenter mentions, the success of Cruise is a good counterexample.

I think you are ignoring the reasonable point and splitting hairs. One might equally say few/none of the US examples would have been possible outside of the US market (where "ask for forgiveness" is not an option), or that Cruise never shipped.

Most deep tech and hardware projects can also benefit from a ship-and-iterate model. There are always pieces you can show to some customer, even if they're simulations or a tiny component of the whole thing. It creates some work and stress to deliver them, but the feedback you get means you're more likely to end up with something the customer wants at the end of the project.

I too embrace your perspective but it is unfortunately naive. If you want a tiny plastic enclosure around a simple circuit board, by all means iterate quickly. When producing larger hardware it is common to have parts which require multi-month production and testing cycles and 5-7 figures USD just for the tooling to enable the process (not including design, actual fabrication equipment, outsourced equipment time or stock material). You usually have multiple parts of this commitment level within a larger product. You don't rush out the door when you are dealing with those sorts of realities.

I try not to appeal to authority, but the parent comment was from tlb, founder of anybots. They’ve been building and selling robots for... 12 years? Probably closer to 18.

It sounds like they might be suggesting that a smaller prototype can be shown to customers. 3D printing in particular can help with that; I think Valve demo’d a new controller that was partly 3D printed several years ago.

The GP was referring to "hardware companies not manufacturing consumer or IOT widgets". Anybots is obviously manufacturing consumer widgets.

There are spaces where even a small prototype will easily run into the six figures in pricing.

What are you building?

The consumer view is always available, meaningfully personalized food produced on demand from fresh ingredients as an autonomous urban utility, 10-20× faster than delivery services. Technically, we're building the robotics, R&D and manufacturing capacities to define and dominate that market with a network of just-in-time contactless kiosk-format microfactories. As a consequence of the challenging form factor we've set for ourselves, the regulatory requirements of food handling and so on, let's just say the complexity and BOM length is arguably greater than most vehicles.

As someone who's had some ancillary involvement in a hardware product I have a lot of sympathy with the disconnect between doing hardware and the "ship fast SAAS product" mindset.

However, in this case your comment:

> As a consequence of the challenging form factor we've set for ourselves

does seem to indicate that decisions you have made yourself are one of the factors causing you to slow down here. Something to consider...

Well, if you survey the space we are looking at you will find good commercial reasons for that form factor. Such as: size of available spaces under typical vending leases, fits through door, fits in lift, fits in corridor, fits in truck, etc.

I don't think anyone will ever argue that you know your market well!

I think what people are trying to say is that there maybe ways to get a partial solution to a smaller part of your customer base much quicker and for less money.

Yes. There may be. But there's no solution cheaper or faster than letting someone else prove the market on your behalf. In our case, we have seen enough proof in the last few years: last mile delivery commodification, 50% of every $1 spent in food retail is convenience-driven, etc. We do not need to prove humans want food or that frustrated demand exists. What we need to prove is that we can supply it, cost effectively, profitably, at scale, with technology we uniquely design, manufacture and operate with a defensible lead.