Breaking orthodoxies in venture capital
My chat with Lee Edwards, General Partner at Root Ventures
Today’s chat is with Lee Edwards, GP at Root Ventures
Sar: Root breaks a bunch of VC orthodoxies. You went all in on hard tech. You have stayed close to your seed roots. All the partners are engineers. No one’s ex-banker or consultant! Your new website is fun and weird. All of this says a lot about Root’s ethos. Root has thrived and gone against the grain in the world of VC, which is a sea of sameness.
Lee: None of us can envision a world where we aren’t building for fun, and that comes down to the areas of engineering where we have worked professionally, as well as the hobbies that tend to be building in one way or another. I built our website; I built our wedding website. Chrissy designed one of our hardware projects and insisted on assembling the boards by hand herself. Emily is learning how to fly. Kane fixes his motorcycle. Avidan jury-rigged the coffee machine and the soda fridge.
I honestly don’t think we’d be happy if we weren’t doing these things. We probably couldn’t spend 40+ hours/week as investors if we didn’t also spend our spare time building things, at least for fun. Most engineers of any discipline can relate to that.
Sar: How does having your specific opinionated way of doing things create blind spots?
Lee: One thing that keeps devtools interesting is the many personas of software developers. I think through my lens, which is more of a web/mobile/full stack Ruby and Javascript sort of person. While I’ve written code since middle school, built plenty of backends, and worked in robotics for my first job, the Ruby on Rails era of programming is my comfort zone.
So the blind spot is the risk of overfitting my model on one user type. Like a consumer investor who only invests in things they would use and misses important trends in Gen Z or other demographics they aren’t a member of.
I have friends and professional contacts who aren’t anything like this. Some hate Ruby and love Java. I know new grads who know the ins and outs of PyTorch but couldn’t build a web front-end to save their life. Some of them even use Windows.
If I didn’t keep building new projects all the time, I’d quickly get to the point where every language, framework, and best practice I used to know is no longer in use. Fortunately, this gives me an excuse to always try new things.
Sar: We don’t have many engineers in the venture business. Given all the talk about investors having empathy with founders at the earliest stages, you would think the most irrelevant talent pool to recruit new investors from would be those who made pitch decks and spreadsheets for 80 hours a week for two years for mature companies. What’s going on?
Lee: Many engineers simply don’t want to give up building for a job that’s mostly talking (and far more sales than I knew I was signing up for.) Most firms don’t value that and won’t give them the space to do it. I’ve had GPs at big firms ask me why I’m wasting my time coding. They believe they can understand the technical founder and the engineering customer through analysis alone, but I doubt it.
I think we are going to see increasingly that returns will be driven by which competitive deals can be done, and at the early stages, at least, founders are going to pick VCs who not only understand what they do but speak their language. It’s hard for me to have conversations with people without using terms like “priors” and “overfit,” which is truly not an affectation. I think that makes my impedance mismatch with technical founders much lower. If that’s right, and this translates into deals won and therefore returns, then I think the trend turns. But given the timescales involved, venture turns more like a cargo ship than a motorboat.
Sar: Community development efforts in the dev tools landscape, overrated or underrated?
Lee: Underrated? But not often done well. Developers have pretty finely tuned bullshit and cringe detectors. There’s nothing better than an organic community that emerges from a place of love for your product.
Sar: What could use more of an engineer’s mindset to make faster progress?
Lee: Running the city of San Francisco. There’s a new example of this every week. Last week, the San Francisco Chronicle interviewed a data scientist the city hired who outlined a process no engineer would design, and no engineering manager would approve. During the months-long process, the city initially told the candidate, who had a PhD in astrophysics and taught a class in data science at UC Berkeley, that her bachelor’s degree did not meet the minimum requirements.
This is some big company bullshit. Any engineer would look at the metrics from this hiring process and make a change. We do this all the time in startups as we apply continuous improvement to our processes, even the non-technical ones. How much do you want to bet this process hasn’t changed in years or perhaps decades?
Sar: What is overblown about the recent enthusiasm around creative AI in the short term?
Lee: AI-generated visual art is mind-blowing, and most of us could not fathom that a computer could do something like this 20 years ago. There are many forms of artwork that people want and value. Things representing thoughts in an interesting person’s head will always be a category. Some sense of ownership of an original will always be a thing. While AI-generated art may have plenty of interesting use cases, like eliminating repetitive work or appreciating the novelty of art generated by a computer, there will always be a market for art made by humans, at the very least in the same way that well-made vintage items hold their value and are highly desirable. The concept of historicity will always be a thing.
Sar: How would you get non-techy normies excited about the potential of computer vision?
Lee: It’s very hard for me to imagine not finding computer vision exciting. I have no idea how to speak to such a person.
Sar: We haven’t made nearly as much progress in measuring software engineering productivity using software as in other functional disciplines. Do you agree? Why is that the case?
Lee: Software engineering productivity is difficult to measure because it is a fundamentally creative endeavor, not a manufacturing process. Unfortunately, that can’t let software engineers off the hook for needing to care about delivery and productivity. Business needs demand an understanding of our productivity, even if just to develop the measures to improve it.
One of the only methods I believe in is Pivotal’s agile/XP estimation process, which uses historical team velocity to attempt to predict future team velocity. It encourages team buy-in on targets and commitments. It encourages personal responsibility and accountability to your teammates through standups. And addresses process problems first before assuming that schedule slips are due to personal failures through the agile retrospective.
A lot of things about agile are not popular anymore, not in the startup world. But other than setting deadlines and asking people to commit to a death march to meet the deadlines, I’ve never seen a better way to predict when a feature will land.
The problems with the alternatives, which mostly measure hours and lines of code, are obvious to any experienced software engineer. Sometimes great progress is made with very few lines of code, and quite often, long hours translate to little progress, while a random breakthrough can reduce a large task into a small one.
Sar: What common pitfalls do technical founders fall into when hiring their first few engineers?
Lee: Primarily whether this employee is a good fit for startups. Most big companies measure and judge employees on all things that have no place in a startup whatsoever while underemphasizing or punishing the things that matter.
This isn’t because they’re stupid. They just operate in a very different context. Most of the company is not seeking product market fit. Individuals need checks and balances on their unilateral decision-making to avoid the risks of any errors they might make. A startup is the opposite of this, and some very talented engineers with perfect performance reviews are constitutionally incapable of operating in an environment where the only thing that matters is results.
Sar: Is there any standard advice that harms more than helps technical cofounders?
Lee: The most harmful advice a technical founder can get at the pre-seed stage is, “You need a business cofounder.” Fortunately, few people are stupid enough to hand out this advice that often anymore.
If this still has to be said, you don’t need a business cofounder. You need to fill the gaps in your business sense. Whether that’s management, finance, fundraising, modeling, sales, or marketing - learn or hire. But there isn’t enough work for one person to spend 60 hours a week in year 1, so no cofounder role exclusively does this in a technical roadmap-driven company. Bring on another technical cofounder instead. This will also free up more time to work on your gaps.
Sar: As a steward of capital, what are you terrified about?
Lee: That 2020-2021 valuations were off the charts, and companies have set the bars too high for themselves, making painful rounds more likely. They traded low dilution at their last round for STRUCTURE in their next round. That’s a horrible tradeoff for them. And for us. Fortunately, we are a heavily reserved fund, which should help us steward capital better than the average seed fund if this nightmare comes to pass.
Sar: What is falling out of favor with the investors, and you still see founder interest around?
Lee: MLOps. Broadly defined. Investors are sick of seeing new companies in this space, but the problem statements every MLOps founder has identified are real, so they keep coming up with new ideas.
Ruby. Full-stack development. Both trending down, to be sure, but the communities are still very active, and from a productivity perspective, they’re hard to beat.
Sar: What’s brewing on your mind about the current venture climate?
Lee: Early stage venture does a pretty damn good job of diligence and vetting founders. For the number of deals that get done, and at stages where there is little to diligence - founder resumes, references, no revenue, no product - there is very little fraud and theft. I bet we hear about almost every case in the news. Even if we were short a few or even a few dozen, this number is astonishingly small, given we hand cash out to teams with dreams.
Sar: What about VC twitter makes you cringe?
Lee: Nothing. VC Twitter is perfect.
Sar: Do you hold any views that will rile up both the vocal progressives and conservatives on Twitter?
Lee: Do I hold any that wouldn’t?
I think most social media companies are pretty good actors. They have an impossible moderation task, and we’ve all been disingenuous with our moderation requests from both the left and right - generally asking the platforms to put their thumbs on our side of the political scales and getting upset when they won’t. And giving them no credit for when they get it right.
When it comes to truly bad things like hate speech, violence, and CSAM, where moderation is entirely uncontroversial to nearly everyone, they do a great job by the numbers.
Most people like these platforms more for family and dogs than politics. We politics-addled brain people are the only ones having this conversation, so we are airplane.gif survivor-biased in our belief that social media = politics. The actual numbers tell a very different story.
Both sides get all of this wrong.
Sar: I found it surprising that the tech press didn’t cover your third fund announcement. What happened?
Lee: Well, we did ask around. But we were told, even by close friends, that venture funds raising $150M are the new seed-stage startup raising $1.5M (i.e., not interesting.)
We launched our website, https://root.vc, a command line interface written in Javascript, chock full of easter eggs for people familiar with Linux. When we launched Fund III, we added an “upgrading” loading bar that made it appear to install dependencies. That went viral on Twitter several times, not always instigated by us. Any tech press outlet would be jealous of the numbers we posted each time it got big.
You couldn’t name many VC websites that elicit emotion or inspire anyone to remember what they found there. VCs know this, of course. Nobody takes their website seriously or has any illusion that it drives the deal flow. But for us, it has. I’ve noticed a dramatic increase in the number of people who know who we are. The website speaks for itself.
Sar: What are we not paying enough attention to and deserve more energy and dollars?
Lee: The obvious answer is climate change. I’d love to do more here. But it would counter our ethos of believing that the best investors in an area know the industry as a technical operator. That’s not me (but it is my fiancée.)
Rust developers? They get a good amount of attention and dollars, but it should be even more.
Sar: There are a bunch of tropes around hardware startups and investing in them. Can you talk about the biggest myths?
(Editor’s note: Lee asked his partners to tackle this last question)
Chrissy: The biggest myth is that root is a hardware-only VC!
1) Hardware is hard.
Short answer: not if you know what you're doing. The important thing to note here is that not all hardware engineers are created equal. There's a real nuance between engineers who can prototype, engineers who can ship, and engineers who can scale. They're totally different skill sets (and the differences are magnified compared to the software world). Too many startups miss that nuance entirely when it comes to hiring. And then they wonder why it's so hard.
2) There are "no takebacks" if you ship bad hardware into the field.
Well, sort of. You'd be amazed by how many serious hardware bugs you can fix with clever firmware workarounds.
3) Hardware is highly capital intensive.
Not necessarily. I can get custom PCBs shipped from China for $0.50/ea. Parts of my 3D printer are effectively free. If you know how to optimize and stay laser-focused on what’s important, getting to a shippable product isn't as capital-intensive as people think. Hardware gets expensive when you have to pay for fancy consulting and design firms, when you obsess over trying to make something look like an Apple product, or when you design for a massive scale from day 1.
4) Hardware is easily commoditized.
Not when 1) you consider the whole user experience around hardware, and 2) you build true “deep tech” hardware products that have a strong technical moat."
Kane: There is no "fire and forget" way to go from design to production, be ready to tackle issues and misunderstandings that come up at every step, and be ready to manage your PD firm/trading co/factory/etc. There are engineers with a lot of experience doing this.
Recent chats:
Empowering millions of deskless frontline workers with Rachael Nemeth, CEO of Opus
Exploring the world of B2B payments with Suzanne Xie, Business lead of B2B Payments at Stripe
A better way to build software, with Karri Saarinen, CEO of Linear
Swapping homes for your next trip with Justine Palefsky, Cofounder of Kindred
Unpacking Crowdtangle's breakup with Facebook with Brandon Silverman, ex-CEO of Crowdtangle
Future of coding in India and on mobile with Amjad Masad, Founder & CEO of Replit
Adobe's transformation with William Allen, an Adobe alum and crypto founder