Upverter: What I got wrong

Zak, Stephen & Michael at MakerFaire 2011

It appears Upverter has finally gone offline. Altium acquired Upverter in 2017. And then Altium was acquired by Renesas this summer. And then I got a note from an Upverter user encouraging me to try to get Upverter back / back online / v2. Which got me noodling… What it would look like to rebuild Upverter today? Was there still a gap? Where does AI fit?

But first… Mistakes were made. We can’t do the maths without some honesty.

2010… native in-browser CAD… the Chrome browser was only 1 year old. Javascript was still fucking awful to build with. Flash was a serious choice for some. Typescript v1 wouldn’t show up for 3 more years. We chose to build with Google Closure transpiled javascript - and I think we got this right.

We quit our jobs. Moved into my parents basement. No salaries. No opex. No capex. Lean. We borrowed a bit of money here and there for AWS and flights. Co-founder Stephen had done ok at his SF startup prior and helped more than he should have. Things changed a bit when we got into YC - we squared the debts, hired our first developers, rented a house to live/work in (founders stayed zero salary). We got this half right: employees costing more than the founders forced us to be ruthless about value - that was the good bit. But ~7 years, living in the office, with no salary, and nothing owed to us, was a crazy move - it paid off - but it wasn’t fair to Stephen, Michael or me. Don’t donate your time to a commercial venture.

Living in the office: Upverter would have died if we didn’t. It felt crazy from the outside - but it was exactly right - and very special. It gave us our culture for free. Working-from-home was implicit. Much harder to bullshit when you’re roommates. Things changed when people moved out. Pretty tricky with a family - we would have needed a much bigger house - and hard to ask anyone to do - but I miss those days. Such clarity and alignment.

PCB / semi / electronics instead of anything else. Possibly an error - but the focus was powerful. I’m envious of the guys at Lucid - we talked a lot back then - they did the general thing, hit a huge audience, monetized sooner, and stayed longer. There’s a lesson about artificially limiting your market - doing the more focused but harder thing - vs the broader, distracted, but ultimately simpler thing. Judging by my current interests I haven’t learned it yet…

Launching in late 2012. This was a massive fuckup. We should have been selling during YC - and we couldn’t - because Upverter was at best a prototype. It wasn’t an image or a deck that got output - it was a manufacturing file, probably wrong, that cost money to fab. That scared us into building so much more than a v1. We should have been much more serious about a v1 before YC. Our v1 was a big lift - and we worked hard to do it fast - but it wasn’t enough - we missed something - it took too long / was too ambitious - and it pinched everything that came after.

Not selling for 20m+ in 2012. It was an error for the company - added years of losses & effort - but it forced growth. It made us better humans. It forced us to make Upverter real. It forced me to learn about business. It forced me to become a better fundraiser. It was the right life move - but the wrong business move - and ironically it felt at the time like I was making the opposite call.

Toronto v San Francisco. This one is mixed. We should have stayed in San Francisco - but it would have cost 10x what we spent to exist in Toronto. I should have raised that money, and made it work. But admittedly I probably couldn’t have. And moving to Toronto allowed Upverter to exist and be a win - where SF probably would have killed it. That said - Toronto made us slower - which also hurt us.

Built-in physics simulation. Mixed. It was exactly the right thing once concierge was in the equation. It was ordered wrong. It was also the wrong word. It was supposed to mean: correctness. But instead it was another sharp edge to cut yourself on.

Parts Concierge. We should have fucking started with it. It was a game changer. Taking full responsibility for the use of our software - the inputs - the ops which were clumsy - not just the code compiling - the fucking fab files and everything that went into them - forced us to get better SO SO FAST. It forced us to be real. It got us taken seriously. It produced real revenue. It aligned our stretch goals (eg simulation). And it unlocked licensing which previously was a dead channel. We ended at the beginning. In hindsight we should have been doing contract CAD the whole journey - and building the tools we needed to do it well.

Mechanical / 3D. Right. But we should have done more. Once concierge was in the equation we had amazing 3D pcb renders - sharable in a browser - collaboratively editable / discussable - before anyone else. We were legitimately ahead of our competitors for a second. This could have been a true mixed-pcb mechanical edge. A niche we could have owned. This could have been (and has since become) a product in its own right for all engineering. But it competed for resources with concierge and simulation - and I probably didn’t invest enough in it.

Meta: Was Upverter a good business? a good market? a good startup? Let’s enumerate:

• Market: Terrible.
• Timing: Early.
• Product: Strong at the end; Took too long (but didn’t cost too much).
• Growth: Mixed; Good relative to invested capital.
• Financing: Minimum viable; below viable if fairer.
• Mission: Directionally right; Too niche.
• People - Cognition: Very strong.
• People - Alignment: Strong.
• People - Culture: Strong.
• Exit: Nominal win for most investors; Good for founders.

We thought we were building CAD software. We should have been doing CAD and building software to make ourselves impossibly good at it.