Our wives are friends. That’s the true answer how GitHub cofounder Tom Preston-Werner and I reconnected over dinner a few months back. But when he shared news of evolving his angel investing into a firm called PWV, I knew we’d have a lot more to chat about. And I’d have at least Five Questions for him….

Hunter Walk: We’re both Gen X, which Google’s AI snippet notes was the ‘first generation to grow up with personal computers and MTV.” I know the former played a role in your life obviously, but what about the latter? Did you watch much MTV growing up? Generally a music fan?

Tom Preston-Werner: I was never as much of a music fan as many of my friends growing up, though I did listen to a certain MC Hammer cassette tape so many times that I wore it out. I guess I enjoyed music, but was never particularly introspective about it. I liked Paula Abdul and Vanilla Ice when they came on the radio, and got roped into buying tons of CDs through BMG’s shady “8 CDs for a penny” promotions. Stuff like Kansas, Stone Temple Pilots, and the Jurassic Park soundtrack. Maybe I was more into music than I remember! But it was never part of my identity. At least until I discovered Pink Floyd in high school.

I did watch a bit of MTV, but I liked the TV shows and commercials better than the music videos. I savored every moment of every Æon Flux episode I managed to stumble upon. That show is deeply seated in my core memory and probably explains some random pathology embedded in my limbic system. I couldn’t believe the weird self-promotional shorts and interstitials they aired. They did stuff on MTV that no other channel would touch, and I loved that. It expanded my perception of what you could do (or were allowed to do) with the medium of television. That same spirit of breaking the rules and expanding what’s possible is exactly what draws me to founders today.

HW: You recently blogged about evolving your angel investing strategy into a proper venture fund (PWV). Your ‘true north’ is to be “the venture firm we wish we’d had early in our startup journeys.” What was your relationship with VC firms like during the GitHub growth?

TPW: For the first four years of GitHub, I actually had a bit of an antagonistic approach to the VC world. We were fully bootstrapped during those years and ran the company entirely off income from GitHub customers. As such, we never approached VCs and didn’t spend much time thinking about them. We took it as a point of pride that we didn’t need to play the VC game to succeed. We made our own rules and were the masters of our own destiny, thank you very much!

The first VC that I recall ever reaching out to us was Mike Maples, Jr. He had put together some crazy algorithm (remember, this is 2008-ish) where he had an associate collecting data about the startup ecosystem and ingested it into his proprietary system and out would pop promising founders. That’s how he said he found us and we chatted with him a bit, but had no interest in taking VC money. I still think it’s cool he put that together way back then.

Even as a VC now, I think if you can bootstrap your company, go for it! When you have to rely on your customers to fund your company, you get very good at focusing on what matters, delivering value to people. This probably requires finding product market fit very early, but if you can do it, it’s a powerful way to control your destiny. At GitHub, we eventually did raise money, but we were able to do it on our own terms (literally). We wanted to go faster and our thin buffer in the bank was limiting our ability to hire rapidly to meet the opportunity in the enterprise. Partnering with Andreessen Horowitz for our $100M series A allowed us to grow faster and access advice around becoming more sophisticated when it came to finance and sales. 

A simpler version of my ‘true north’ is that I want to help founders succeed. Not just because I want to plug a big number into a financial spreadsheet down the road, but because I really BELIEVE in founders and meeting them where they are. I want to dig into the technical details and ideate on the product. I want to help them make the thousand small decisions that prepare them for smooth sailing later on. I want to discuss branding and hiring and how to build an exciting, sustainable culture. I want to feel less like a VC and more like a mentor and champion.

HW: Raising a fund is one task – often easier than actually running and deploying it! Has anything changed about why/how you invest when it switches over from your own angel dollars to a proper fund structure (which includes outside capital)?

TPW: Angel investing is great, and I love doing it. But it means almost always being a small player on a big cap table. When David, DT, and I joined forces in 2022 to start the Preston-Werner Ventures Rolling Fund (a quarterly venture vehicle administered on AngelList) we brought in some outside LPs and drastically enhanced our evaluation and diligence process. We started writing a few bigger checks, and in doing so, found that many of our portfolio companies reached out to us when they were ready to raise their next round, asking if we could lead. But with a small fund, we had to say no. And that felt like a big lost opportunity to be more involved and help these founders even more.

By raising PWV Fund I, we’re putting together a vehicle that extends our experience in funding a broad range of startups, and adds on the ability to double down on our most promising companies. We’ll more often be able to lead rounds and set terms. It’s really about leverage and opportunity. In addition, we’re selecting LPs that want to join our community and help out. When we can leverage the networks of a broader group of investors, we can really move the needle.

Some things won’t change. For me this is much more than a financial game. I want to fund founders that want to create a better future for PEOPLE. Which means we spend a lot of time thinking about the ethical ramifications of each investment and how the technologies they enable will impact people’s lives in a few years, should they succeed. In speaking with LPs, we look for like-minded people that also want more than just a financial win. As a capital allocator, what we really are is a TIME allocator. And there is nothing more sacred to me than ensuring that the people I influence are working on meaningful, human-centered, value-creating technologies.

HW: SF is in a doomloop. SF is back. Rinse and repeat for those of us who have been here a while. Certainly a special place, but also a city with challenges. If a founder asks you whether they should base their company here versus somewhere else, what’s your advice to them? When you invest do you care about company location?

TPW: The Bay Area is awesome for startups. Everything here is optimized for founders working on big ideas. VC money is headquartered here. Legal folks know how to manage cap tables and stock grants. Banks understand that startups are different from brick and mortar establishments. If you’re building a startup in SF, you won’t be alone. There’s an AI event practically every night. Stop by a random cafe for a flat white and you’ll hear a founder pitching their idea to an investor or candidate. If you can find a way to live in the Bay Area, it will absolutely accelerate your timeline and create opportunities you won’t have elsewhere.

That said, there are downsides. You will spend a lot of money to be here. SF salaries are often four times what you might pay for similar talent in Germany or the UK. That means your runway is accordingly shorter. So you have to raise more money, which means more dilution. Maybe the upsides make all of this a moot point, but maybe they don’t and that extra year of runway could allow you to find success. One thing is clear, though. If you are building in SF, you better burn twice as bright, because you might just be burning half as long.

At PWV, we’re happy to invest in companies outside the Bay Area, because there is talent everywhere, and good ideas don’t care about geographic boundaries. But if you do start your venture somewhere else, you need to deliberately make up for the downsides. It means you need to make relationships with VCs and others in SF so we can help you get hooked into the right services. If you’re outside the USA, it’s especially important to have partners (VC or otherwise) in the USA to help you incorporate in a way that makes future funding rounds easy for US investors and smooth for US customers to pay you. We do all these things, and it allows us to help founders wherever they may be on the planet.

HW: Who is a person in tech – dead or alive – that you think deserves more recognition than they receive for their impact upon our industry?

TPW: My answer is not a single person, it’s a KIND of person. It’s every person that has ever contributed to an open source project because they believe that when we work together and share our best accomplishments, we raise the bar for what we can do as an industry.

Untold thousands of developers have spent time writing and giving away code, fixing bugs in critical open source infrastructure, and fielding support requests from users, all without being paid or recognized as the heroes they are.

You will know the names of very few of these people, yet their code makes our modern world of technology possible. Nearly every device you use, from your car to your phone to your watch, and certainly any software product you build and deploy will run and rely on open source libraries. A single line of code written by an open source contributor could run trillions of times a year on various computers around the world, all without that developer ever earning a dollar for it.

And that’s ok. It speaks to the belief so many of us have in the idea of giving back. Giving back to the community that helped us achieve what we have by standing on the shoulders of open source giants. But it also means we should think about and be thankful to open source authors. It means that when we submit an issue on an open source project, we should start with a big “thank you” and end with the best damn issue report the maintainers have ever seen. We owe them that. In fact, we owe them the very success of our industry.

Thanks TPW! Hope we’ll have lots of founders to back together! – HW

Junova tracks your flights and requests credit for you when prices decrease. A friend started this as a ‘lifestyle business’ – i’m giving it a whirl. It’s free – just take a percentage of the refund. Sign up and get $25 credit to start.

Digging in the crates at home I found this from Karen Kavett (who is now doing puzzle videos but at the time was more DIY stuff). It’s a fantastic snapshot of an era, I’m guessing 2010/11/12? I wonder what it would look like today now that everyone realizes that YouTube is for, well, anything and everything!

Junova tracks your flights and requests credit for you when prices decrease. Friend started this as a ‘lifestyle business’ – i’m giving it a whirl. It’s free – just take a percentage of the refund. Sign up and get $25 credit to start.

I like words. Here are thousands of them.

The Boss Who Gave His Employees a $240 Million Gift [Gregory Zuckerman/Wall Street Journal] – Family owned business in Louisiana sells for $1.7 billion and the owners negotiated that 15% would go to 540 full time employees. Employee ownership is such a given in our industry but doesn’t extend to most jobs. Tech comps also tend to distort our perspective of outcomes since power laws can result in millions and millions, but for most people a six digit bonus is meaningful and life changing.

Sequencing vs. Equal Odds [Michael Dempsey/Compound] – I usually only understand 75% of what Michael writes, but this one resonated soundly enough to share! He outlines a theory around two types of companies/R&D: those which require sequencing through the idea maze and those which are better suited towards parallel experiments because you don’t know which is going to work best. There are successful outcomes available to either, but require very different skillsets and strategies. To be between the two is death.

Pre, Mid, Post-Training Way of Life [Tina He/Fakepixels] – Another banger from one of my favorite writers/chroniclers of the community right now. Here she applies (extends?) LLM metaphors into life and everyday language. Three different mindsets out and about in the world; the trio in the title. “Token by token, brick by brick, we train ourselves either toward a larger freedom or toward a more elegant cage. And the difference is rarely intellect. It is what we are willing to protect as sacred.”

Soft Serve Superpower [Dave Margulius/Electrified] – Dave’s an ex-operator (cofounder of Quizlet), who now works on climate issues. Soft Serve Superpower reminds me of Noah Smith’s Electric Tech Stack thesis, but tells the story via a surprising vessel: the American ice cream truck. And connects it to the front lines of Ukraine’s war with Russia.

What We’ve Learned About AI Readiness in Real Estate Today [Brad Hargreaves, Jonathan Gheller/Thesis Driven] – Sorry it’s behind a paywall but I wanted to at least put it down here because its main point is something I agree with pretty strongly for almost every business vertical today: namely that AI rollouts in industry are increasingly gated by security, governance and observability/reliability, not skepticism of the technology’s ability to create value. It’s why I’m bullish on AI but believe we’re going to be looking at 3-5 years of ‘rollout’ in many industries, with early adopters or AI-native players getting a headstart.

Enjoy!

Junova tracks your flights and requests credit for you when prices decrease. Friend started this as a ‘lifestyle business’ – i’m giving it a whirl. It’s free – just take a percentage of the refund. Sign up and get $25 credit to start.

I got to know Alex Heath during an earlier incarnation – maybe his Business Insider days? But always knew his entrepreneurial streak would lead him to do his own thing at some point, and now he has! Sources.news is his tech publication, covering “what’s next for AI and the tech industry.” Worth subscribing! I wanted to know more about this transition so here’s Five Questions with Alex.

Hunter Walk: You recently started Sources, which you describe as covering “what’s next for AI and the tech industry, featuring original reporting and unique access to the companies at the forefront of the AI race.” Why is this a beat you thought best done indie as opposed to inside an existing entity? Or was the motivation something different?

Alex Heath: I went indie for two simple reasons: it’s something I’ve been wanting to do for a few years, and everyone I know who has done it is glad they did. AI is the biggest story in tech ever (and maybe the world?), and it’s an area I’ve been steadily increasing my focus on over the last few years. It’s not so much that I thought the beat was best done on my own, but it is freeing to cover these crazy times in a way that is totally authentic to my interests and style.

With the podcast, Ellis Hamburger and I want ACCESS to feel more like your favorite tech interview/talk show. We sometimes talk about ‘Smartless for tech,’ but we are only a few months in and still figuring out the vibe. We’re not dumbing down topics for a mass audience, but we always want it to feel entertaining. I see ACCESS as part of the same cinematic universe as Sources, but it’s where I get to be less of the ‘reporter’ and just hang out more.

HW: From the folks who’ve gone Indie before you, who do you particularly admire and what is Sources borrowing/evolving from them?

AH: Our mutual friend Casey Newton has been a huge inspiration and has pushed me to do this for a long time. I really admire how Emily Sunderberg is 100% herself and such a natural at taking advantage everything that makes the Substack ecosystem unique. Stylistically, I’m probably closest to Matt Beloni. If I could build something half as relevant to the tech industry as what he has built for Hollywood, I’ll consider Sources to be a success.

HW: When a source shares news with you, how do you decipher what their motivation might be, and how does your estimation of their trustworthiness factor into how you use that information?

AH: It’s pretty easy to tell this upfront. Usually, people share information they shouldn’t because they feel slighted, want to force change, or just like to gossip. Often, I find that the people on the other side of leaks discount the prevalence of that third reason. Motivation doesn’t really impact trustworthiness if I’m getting verifiable facts (internal docs, chat logs, etc.), but I always try to corroborate and assume the worst intentions from the start when engaging with a new source. There’s probably nothing that bothers me more than getting something wrong.

HW: An aspiring new grad tech reporter asks you for one piece of career advice – what would you tell them?

AH: Do your damndest to become a world top-three expert in a specific, valuable niche of tech reporting. I spent about a decade trying to be the best social media reporter in the world, and not to toot my own horn, but as far as I can tell, I have more scoops under my belt on Meta and the other key social media companies than anyone else. For awhile, I knew Facebook’s org chart better than a lot of people I talked to who worked there. That’s the kind of obessive detail you have to be willing to drill into as a baby reporter. 

Right now, if I were a new to this, I would try to become the single greatest OpenAI or Anthropic reporter in the world and manically cover the ins and outs of those two companies. If you start breaking news there and can be consistent, your career will be set.

HW: Looking back over the last few years, what’s a person, company or trend that you were sure was going to be really important and ended up being less impactful? On the other hand, what’s something that you largely ignored but then became too big to not cover (or maybe you just got religion late)?

AH: While it’s more relevant now, I certainly thought augmented reality glasses would be a bigger deal than they are today. I covered AR and VR a lot over the last decade based on the investments and hopes the leading players in the space were projecting. It’s taken a lot longer to hit as a category than everyone thought, though I do think this will be an interesting year with new Meta glasses, the return of Google Glass, and Snap’s consumer Specs glasses coming out. AI is also making the form factor more compelling, which dovetails nicely with my past and current reporting focus. 

I deeply regret not paying more attention to AI as a field before the release of ChatGPT. Like everyone, I’m certainly paying attention now!

Thanks Alex! Folks should check out Sources!!!

New Year’s Links

Backchanneling is Becoming a Crutch [Cristina Cordova/Linear] – I love a good backchannel reference but it takes skill and knowing what you’re trying to accomplish in order to get the most from this activity. Here Cristina points out how you can make mistakes using this tactic – especially in exec hiring. “If you want executives who can actually move a company, you have to do the harder work yourself: understand the role, accept the tradeoffs, and talk to the person before letting someone else’s context make the call.”

How the Hell Are You Supposed to Have a Career in Tech in 2026? [Anil Dash] – I do sometimes wonder if I was 20 years younger how I’d be approaching my career right now Would AI be exciting or freak me out as an existential risk? Would the culture of tech be as attractive as it once was now that it’s powerful, not the underdog? Would I find my tribe?

Strange Math [Ashritha Karuturi/Ambrook] – Ambrook makes financial software for farmers and is a company we’ve backed since their first fundraise. The team is so smart, so intentional, so iterative in their willingness to learn and adapt. Here they write about tactics for building a culture of radical honesty.

How Kraft Heinz Lost Its Lock on Mac and Cheese—and American Shoppers [Jesse Newman/WSJ] – They’ve given up mindshare and shelf space to new competitors.

The 9s of AI Reliability [Kushal Chakrabarti/Obviously Wrong] – Opendoor’s former Chief Research Officer, and a AI/ML advisor to startups, takes on the question of ‘how reliable does AI need to be to be of value’ and notes the answer is quite different depending on the use case and situation.

Two Things to Sign Up for in January 2026

1. Founders Card: If you travel a lot for business, it’s worth it alone for the discounts on hotels, airfare, etc. Generally a perks-based membership club. I’m usually dubious of these but have done this for several years and always renew.
2. Blackbird.xyz: Local app [SF, LA, NYC, Denver] – get $20 for free on first ‘check in’ – coffee, bars, restaurants. I’m a former 4SQ/Swarm users and trying to figure out if this is interesting or not. But in the meantime, you’ll at least get some venture capital subsidy 😉

Most Silicon Valley folks expect some degree of ‘tech talk’ when returning home for the holidays. Sometimes it’s in the form of IT support. Or nephews excited about the latest degen trends. The Uncle who is very excited to share he cut the cord on cable this summer and went all in on YouTube TV. But last week there was a single topic of conversation: are we in an AI Bubble? I expected this from portions of family tree, but it came from, well, everyone. My mother-in-law wanted to know if she could visit a Data Center with her friends. Basically my holiday table was no different than a San Francisco Blue Bottle.

Despite the Bubble question being topic du jour back home as well, I was kinda unprepared for how to summarize my current feelings. Largely because regardless of the outcome, our industry is incentivized heavily to play the game right now. And while I don’t mind shrugging and saying ‘I can argue both sides’ over cocktails at The Battery, my relatives demand more specific responses!

So here’s my 100 Hours to Go in 2025 summary. There’s not one type of bubble to assess, but instead four.

Technology Bubble -> NO. You can be a believer in Artificial Intelligence without having to assume it can achieve AGI or Superintelligence. And that’s where I am. The opportunity and promise of the technologies currently being built is as impactful as every other phase change I’ve lived through: personal computer, Internet/WWW, cloud computing, and the smartphone. I’m incredibly excited about both edges of applications – the mundane and the advanced – while thinking the middle is going to be harder to solve. What I mean by this is, lots of small productivity gains doing basic repetitive tasks and hyper-compression of complex simulations and analysis, but workflows that require multiple points of offline interactions or situational judgment or five 9s accuracy, are still elusive. That’s fine! The small stuff actually adds up to a TON of productivity gains. And the large stuff (drug discovery!) is hugely impactful.

Investment Bubble -> Likely YES. By this I mean, will the total amount of capital being deployed into the AI space provide in aggregate its expected returns (whereas the Valuation question below is the price of each asset). Here I tend to buy more of the ‘telecom buildout’ bubble thesis – that it’s not clear the starry-eyed modeling for $2 trillion more of data center capacity and so on is going to have the IRR that’s modeled today – the same can be same for chips that could lose value faster, and so on. But that these assets will still be *productive* assets over an extended period of time, not nothingburgers. So capital markets, circular business development deals, and the like are supporting an over-investment, that will decrease overall returns of much of the category, but that the yields will still be positive overall (yes, some players will go to zero).

Valuation Bubble -> Absolutely YES, in the private markets. I can’t even really bring myself to write much here. Read Pace Capital’s Chris Paik’s “Paying $3 for a Dollar: The Rational Irrationality of Venture Capital.”

Revenue Bubble -> NO. If you read my Technology Bubble paragraph above then this shouldn’t be a surprise. We could stop new model development and current model improvement RIGHT NOW and I believe it would still take 3-5 years for society to understand how to adopt all the use cases the current stack supports. Because it’s not just the input/output part, it’s also the changing of processes, strategies, and organizations to realized all the benefit of these tools. There are various reports which say “AI Revenue Must Grow XXX% YoY” to fulfill current expectations, and folks often repeat these numbers with skepticism. I’m on the bull side here – the pie is going to grow much larger over time. This isn’t just about company pilots and consumer experimentation (although there will be many shifts in where the revenue goes over the years), this is evergreen.

So I guess I’m generally bullish on the technology and business, and believe there’s a bit of a mania right now in the funding, which might help society overall but is a crazy time to live through.

Boxing Day Link Blog!

The Rumor Mill and the Propaganda Machine [Renée DiResta/Agents of Influence] – I never pass up an IRL opportunity with Renee, and we grabbed coffee this past summer after finding ourselves in the same city during family travel. Goodness she is smart and strong I thought leaving the conversation. In this post Renee puts context around the X mob that named and harassed a Brown college student they suspected of being the mass murderer, in which he was not at all implicated.

I’m not on social media much and heard about this incident only because horrifically some powerful folks in the tech community participated (most later deleting their posts). My first reaction was that this was closer to yelling fire in a crowded theater than freedom of speech. Renée discusses how reality has splintered and why consequences have decreased as well. She also touches an interesting third rail:

It’s my controversial opinion that defamation lawsuits should be far, far easier for private individuals who get screwed like this. I just don’t think the law has caught up to the infrastructure. Yes, defamation lawsuits are imperfect, slow, and expensive, and can be frivolously abused. But influencers who falsely accuse a private person of mass murder for clout and profit should face consequences. The legal system is one of the few mechanisms available to impose costs and deter this kind of behavior. The harm is real. Treating human beings as content is depraved. Defamation has always been outside the bounds of free speech, and our marketplace of ideas would benefit from less of it.

Year 5/6 Startup Learnings [Celine Halioua/CEO of Loyal] – Always love builders who periodically take a step back to share. In this case, Celine’s annual Learnings posts. The fifth of these posts includes several notes to chew on. My favorite is “You don’t know someone until they make a mistake.”

Self-driving cars are an unambiguous social good [Mathew Ingram/The Torment Nexus] – 2025 seemed like a big tipping point for Waymo going mainstream and the narrative of an autonomous future shifting another standard deviation in the positive direction. Heck yeah! Mathew writes a great roundup of the benefits here. Of course I agree! I was at Google during their earliest years of work in this space, enthusiastically reading whatever I could about the team’s findings! Then we invested in Cruise’s seed round and got to see another approach starting to work. In 2017 I worried that our measure of ‘safety’ for autonomy would be illogical and emotional. In 2023 I talked about how these might be the most inspirational use of technology in physical spaces that we’ve seen in years. And in 2024, I urged regulators to let more people die to move this space forward.

Dashboards or Pipes? [Gokul Rajaram/via LinkedIn] – Gokul was our colleague at Google back in the day, and in this specific context, our co-investor in Graphite, which was acquired by Cursor earlier this month. The outcome made Gokul recall, and resurface, some advice he’d given the Graphite cofounders about his theory on Dashboards or Pipes. Short post for you to read but the tldr is Dashboard products are used directly and regularly by end users as their primary interface for accomplishing tasks while Pipes products are used in the background, and you need to figure out strategically what you are.

Episode #60: 2025 in Review [The Learning Corner podcast/Precursor Ventures] – For me podcasts are usually escapism – NBA, history, some pop culture, maybe some 1980s pro wrestling. I very rarely listen to tech or tech-adjacent discussion but this pod has always been one that makes it into my queue. They take ~three articles that their firm discussed internally that week and share their thoughts. It also helps that I often agree emphatically with Charles Hudson (we’ve known one another for a very long time) and find that co-host Mia Farnham adds much to the mix. The one I’m sharing here had their first guests – Me! and Peter Walker from Carta.

AND OMG, adding this link I see that I spoke for 43% of the pod. Sorry, not sorry!

Enjoy!

I always love sharing something that delights me. My favorite purchase of 2025 might just be these microwavable bowls that are excellent for instant oatmeal.

Org charts are more like funhouse mirrors than many people realize, providing a view of the truth but with multiple distortions. If you assume that all leadership is hierarchical, or everyone with the same title has the same influence, or that the number of people you manage is directly correlated with your importance, well, the way it works in Boardrooms, corporate offices, and hallways after hours can be dramatically different. During my days at Google there were multiple examples of “Eric, Larry and Sergey” whisperers who could help bless or kill a project without having a SVP title. There were senior engineers scattered about without any teams but who could look at your proposal and tell you whether the architecture would work or not. And there were culture carriers who served as weathervanes for morale, as well as being weather makers themselves.

These lessons never left me – I experienced them positively and sometimes less productively as a product exec there. Working with startups today via our venture firm Homebrew, they inform the guidance provided to the founders as they build and scale their own companies. And equally I try to remind people within those teams, especially the ones that are so impatient to get to a certain title or headcount responsibility that they can be their own worst enemies making career decisions, that Span of Influence is Not Limited to Span of Control. Don’t run to a worse startup just because you can get a VP role there, or OTOH stay somewhere uninteresting out of concern you’ll have to ‘take a step backwards’ [title/managerial-wise] applying for other positions. Again referencing my Google years, there were a ton of early business and operations team members who despite their seniority, came on in IC or lower-titled management roles after understanding the opportunity. You know what happened? They succeeded wildly and got promoted quickly as we scaled, ultimately with a much more satisfying career arc, and definitely more lucrative, than if they had said, nah, that’s beneath me.

As an aside, Mayor Lurie here in San Francisco has been an excellent positive version of this during his first year on the job compared to what I was experiencing on the ground prior. The Mayor’s office does have structural challenges given the Board of Supervisors governance responsibilities and all the various Commissions. GrowSF has written why this is a negative for our city, especially in light of the work to be done in public safety, housing development, and rezoning. Despite that reality, Lurie seems to be an energetic and present problem-solver, giving off vibes of accountability and effort, even if the solutions aren’t fully under his office’s authority. We’ll see if this hopefully can be precursor to collective action and sustainable progress.

Some of you should skip this post. It falls into the ‘VC reminiscing about time with a startup after a successful outcome was announced,’ specifically software quality startup Graphite being acquired by code generation startup Cursor. Don’t get me wrong – it’s definitely better written and less aggrandizing than the most traditional versions (you’re not going to hear how ‘we’ built this company together), but still, it’s in this category of essays. I just don’t want anyone to be surprised – like when there’s a cute guinea pig on IG and you forget for a minute that they’re really just rats with a blowout.

With that warning done, yes, we led the seed round for Graphite in early 2020. I met Tomas first via a mutual, and then all five of us (me, Satya and the three Graphite cofounders) all did some quick conversations in various combinations. Satya’s feedback to me was memorable – something along the lines of ‘please get them a termsheet before other people realize how smart they are.’ And we did. It was a wonderful almost six years together and I know the relationships are going to outlast the transaction. I learned a ton from them about building in the developer ecosystem and from my fellow Board members, Peter Levine at a16z and Christine Esserman at Accel. While there were lots of microlessons, three particular memories are most personally indelible and I’d say if any one of these hadn’t happened the company would not have become what it did.

Took the Punches and Iterated to Durable PMF

Graphite started out as Screenplay. And while ‘code quality’ was always a true north mission statement for them – the reason the company was founded – it took some reps to find the best product manifestation of this goal. These weren’t pivots, but they did require first principles thinking, some intuition, and the confidence to leave something good behind in the search for great. In particular, before what we know today as Graphite was launched they had a mobile QA and release management tool deployed with customers. Given the frothy funding market at the time and the quality of this team they absolutely could have ‘forced’ this version of the company to a Series A. That is, an adequate set of performance curves, a bigger story, and pedigreed cofounders would be enough for a next funding round, and then with that capital they could hopefully figure it out from there. But Greg, Merrill and Tomas knew this wasn’t the way the best version of Graphite was going to be constructed. So they shut down that effort and repurposed some of the code – and a lot of the scars – into a new direction.

As mentioned earlier, going into the seed funding we had ZERO questions about the cofounders intelligence, ceiling, or relationship with one another. We did spend time – with them directly and via backchannels – to see if they could take a punch. If their first idea didn’t work would they be so shell-shocked as to run back to a BigCompany with their tails between their legs? Would the threat of failure be motivating or send them into denial? Would they have enough of a framework or guts to look at something as say ‘this can be ok, but we can do better?’ They did.

We still stress that seed is not a funding round, but a phase. And that most importantly it’s a time for experimentation and learning, not just brute-forcing an up and to the right curve. Or even worse, mindlessly doing what you put in a pitch deck just because that’s what you thought a year ago. The combination of bigger rounds and new/nervous VCs often asks founders to speedrun this process, but the best companies operate at the intersection of urgency and patience. Graphite threaded this needle.

Started During COVID and Back to IRL Quickly

Our conversations with Graphite started right before COVID 2020 lockdowns. It was a very weird time. I had at least met Tomas IRL but otherwise things became Zoom relationships quite quickly. As three close friends the cofounders were able to spend time together in NYC and began to assemble a small team virtually but not remotely, but still there was eagerness to get back to IRL ASAP. And they did as soon as it was advisable.

During this period I got to meet – mostly over Zoom, or in person outdoors – about the first ~20 hires they made. Often as part of the ‘close’ process once an offer had been made but sometimes earlier. It felt really wonderful to be included in this manner – the Graphite team appreciated the help (it was at their request) but more importantly it let me get a feel for the company as it started to grow. So many other startups at that time were sort of black boxes. Who knew where people were with their fake zoom backgrounds? Were people really committed to the cause or just working a few hours a day while trading NFTs and ICOs? This wasn’t the case with Graphite. Getting back IRL was about knitting a culture early and building the trust to do the aforementioned iteration. Put together everyone’s ideas and opinions together.

Were there some candidates who decided to not interview at Graphite because they preferred a remote lifestyle? Sure. But there were equal numbers who actually were leaving their jobs because their current employer was staying URL first. Remember, this was a young tech-focused startup in NYC – working from home with your two roommates and a small apartment is not glamorous. Having the social and physical separation to walk, bike or take the subway a few stops to an office was actually a feature, not a bug, for many. Graphite stayed true to who they wanted to be and ripped off the ‘what should we do’ bandaid early. It paid off – I don’t believe they would have been nearly as successful if they had stayed remote.

Three Founders Went the Distance Together

Point blank: By Year Six it’s unusual to see three technical cofounders all remain vital to the company. In my experience one of two situations tend to occur – they realize the CEO isn’t scaling and try to make that change or the two non-CEO cofounders overlap too much in engineering/don’t scale as leaders. There are versions of this which are polite and productive and versions of this which are toxic and disruptive. But it happens a lot. Obviously it didn’t here. But why?

Merrill wanted to be a great CEO not just a great cofounder. From early on he had the self-awareness that this would take work. And that the role would sometimes put him in disagreement with his cofounders on strategic decisions. He simultaneously embraced the responsibility while also prioritizing the effort to ensure the three of them were always communicating well, trusting each other, and putting Graphite ahead of any one of their feelings or ego.

Greg and Tomas were two sides of the same coin. They both cared deeply about Graphite being a place where engineers could make a difference. To oversimplify, Greg thought about this from a collective and structural perspective (what does the Engineering Org look like, what’s appropriate documentation and processes for our scale and culture, when do we need Engineering Management) and Tomas biased towards the individual (how could any one engineering hire make a difference, how could he look a potential team member in the eyes and sincerely say Graphite would be somewhere that they could have a great idea, build it, and ship it without five levels of approvals and a chilling effect of low risk tolerance), This is of course an abstraction – but they scaled beautifully. Tomas could work on experiments, Growth projects, lead Product within the technical and cross-functional org, while Greg had a natural role as CTO.

Of course I lived on the cap table, not the org chart, so my observations here are incomplete and reflect an outsider’s perspective, but I have such tremendous respect and admiration, not just for what they built, but how they built it. And that’s what we strive for at Homebrew. The intersection of joyful work with great teams and financial outcomes which make the work possible ongoing.

Onward!!!