Code Climate Velocity 2.0

About a year ago Code Climate announced its Velocity offering which provides engineering insights. Today they are taking the wraps off Version 2.0 for everyone. Existing customers have already had access to many of the new features. I am writing about this today because the whole history of the Velocity offering is illustrative of the right way to launch a new product even when you are a company that already has a mature offering (Code Climate Quality).

Many organizations once they have a mature product fall into a trap. They can’t release something new because they can’t reset the dial to behaving like a startup. Instead of going out with an MVP and learning from customers they want to release something that matches the maturity of their other offerings. They often wind up not shipping at all or after a new startup has already come in and taken the field.

Code Climate did a terrific job avoiding this trap. Their initial release of Velocity was intentionally minimal in the best sense of the word (think first iPhone). It was no frills but provided meaningful value and many customers adopted it. The subsequent feature development has been guided by the needs of these customers culminating in today's release.

If you are looking to enable your engineering team and processes with better insights and analytics, go and check out all the goodness that is now part of Code Climate Velocity 2.0.

Marley Spoon

One of the most basic human needs is the need for food. We all need to eat and drink to survive. And so it is not surprising that substantial parts of our lives, economies and societies are devoted to growing, distributing and preparing food.

Food is also a cultural touchstone. We associate certain foods and flavors with geographic regions. We speak of “comfort food.” And there is ample evidence that eating meals together is a crucial bonding experience among friends, colleagues and even strangers. There is even a saying “Families that eat together, stay together.”

It should therefore not be surprising that food is an important area for innovation. And there is a ton of that happening at the moment, from fundamental research on plant biology to synthetic meats to indoor farming. Another area for innovation is the food supply chain, which at present is incredibly wasteful. Between restaurants, grocery stores and fridges in peoples homes, up to 50% of all food grown on farms goes to waste.

As we think about a planet that is headed for 11 Billion people, it is crucial that we work on making quality food and shared meal experiences accessible to as many people as possible. That is not just a challenge for developing nations but right here in the US healthy meals at an affordable cost are hard to come by for many.

Meal kits are one of several innovations that will broaden access. By knowing their customers’ tastes and having pre-orders, meal kit companies can dramatically reduce food waste. Some people will be quick to point to packaging waste but even that can be reduced by meal kits compared to most grocery stores through compostable materials and by circulating packaging (where boxes are re-used).

The meal kit sector has taken a beating because some companies struggled against outsized growth expectations. But that doesn’t change the fundamentals. Meal kits are not for everyone, but they work well for families that are time constrained yet want to eat a home cooked meal.

We are excited to be backing Fabian and the team at Marley Spoon, who are pioneers in the sector. They have led a number of important innovations, including a big expansion of the available menu choices -- now 20 recipes per week and growing -- and the introduction of a more affordable option with Dinnerly, which costs only $5 per portion. Still it is early days and much remains to be done to unlock the potential of this new food supply chain.

The Distributed Computing Update #3

Last summer, I wrote an overview of distributed computing projects in which I now believe I got one key point very wrong.

At the time, I believed that a distributed compute project’s core strategy should be to undercut centralized compute providers on price. At the time, I thought developers wouldn’t choose a harder-to-use distributed compute platform unless it was significantly cheaper than its centralized competitors.

I no longer think that price is the most compelling way for distributed compute projects to compete against centralized cloud providers; I now believe that it’s more about beating centralized providers on developer integrity.


Like Nick described in a blog post last month, when developers invest their time building apps on a platform, they are trusting the platform to be around and available, and not to turn off access to important APIs and features.

Developers inherently understand that building on closed platforms means taking on the risk that the platform will one day shut down or change its rules. There are 14,600 questions on Stack Overflow mentioning vendor lock-in. Developers are hyper-aware of it.

For the first time ever, blockchains make it possible for developer platforms to be around forever (or as long as a single network node is running), and to do so without there being a single company that can unilaterally make changes to the rules of the platform.

Developers can trust that a distributed compute protocol won’t go away, fall down, lock them in, revoke their access, or become their direct competitor. Developers cannot trust Amazon, Google or Microsoft to promise them the same.

This may not be the first time open developer infrastructure wins over closed infrastructure. Brad argues that the reason why Microsoft missed the web was because they missed building an open source server. In the nineties, developers chose open source Linux over proprietary Microsoft.

What about the developer experience

A small portion of developers may care so much about building apps on high-integrity platforms that they may be willing to forgo the great developer experience they’d get with a centralized cloud provider.

We can see this happening already with the developers choosing to build apps where the core logic runs on Ethereum instead of on a centralized platform. It can be a more challenging experience to build apps on Ethereum because the dev tools and libraries are still new, yet some developers are willing to build on it anyway because they value that Ethereum cannot change its rules on them the way that a centralized platform could.

As a project keeps improving its developer experience over time, its developer market share should grow, the same way that more developers started moving to building on Ethereum as there started to be tools like Truffle and Infura that made it easier to do so.

The core team doesn’t always have to be the ones that make the developer experience better. Sometimes those improvements come from tooling built by the community such as Heroku and Zeit building easier interfaces for deploying to AWS, or Infura and The Graph building easier interfaces for interacting with Ethereum. But still, the same principle applies that as it gets easier to use a computing platform, more developers may use it.

The important bit here is that while distributed compute projects can improve their developer experience until they catch up to the centralized providers, it is potentially impossible for centralized compute providers to improve their integrity to catch up to their distributed competitors. No matter what they do, unless they rebuild their platforms on a decentralized network, they will always have the problem that they are a centralized company running on centrally controlled infrastructure. This is just like how Ethereum may after many years achieve a comparable developer experience to AWS, but AWS cannot catch up with Ethereum on developer integrity because it will always be centrally controlled:

Not just compute!

This strategy may work for other developer services beyond compute. FileCoin may be able to offer more trustworthy storage than S3 ever will. Blockstack may be able to offer more trustworthy hosting than Google App Engine ever will. Similarly, Piccolo for databases, Handshake for certificate authorities, and CacheCash for caching. These are just a few examples but there are many. All of these projects offer more challenging developer experiences today than their centralized competitors, but as they improve over time, more developers may switch over from the lower-integrity centralized solutions:

The Opportunity

Nick calls this the startup-sized integrity gap: what centralized players leave lacking in developer trust is a startup-sized opportunity for a decentralized player to fill.

There seems to be four important pieces to successfully executing this strategy:

The first is to emphasize developer trust and integrity by making the platform governance rules transparent and clear. This transparency around changes and governance seems like a critical piece of cementing trust.

The second piece seems to be time - a project has to stay afloat long enough that it can take years to catch up on the developer experience. There are probably two ingredients to doing this: one is to play in a big enough market so that capturing even a fraction of a percent of the market is meaningful revenue, and the other is to be so capital efficient that even if revenue is low, costs are down enough that the team has runway.

The third is to continuously invest in improving the developer experience, and to almost singularly focus on that rather than branching out and building new products and services. This can be done through efforts in the core team, or by working with the community to develop tools.

The final is to consider that while a project may not need to compete against AWS on developer experience on day one, it probably does want to be leading the curve in developer experience amongst the other distributed compute projects.

Distributed compute platforms, by the nature of being built on crypto, can promise developers a level of integrity that no centralized cloud platform can. That feels like a big opportunity.

Ten-thousand better decisions: Building a network around "network effects"

"Normally I’m against big things. I think the world is going to be saved by millions of small things" -- Pete Seeger

There are a million decisions that go into the process of building a company. By surfacing perspectives from 75 companies in our portfolio network, we hope to create tools that help people within each company answer even a fraction of these. Added up across this 10,000-person community, if everyone is enabled to make one better call, then this might lead to 10,000 better decisions than before.

When starting a business, the founding team must decide early on what to create, who it's for, and how to build it. Following that, a new set of decisions emerges, including choices like how to launch, who to hire, and how much money to raise. And so, the process repeats itself.

Albert Wenger, one of our partners at USV, likes to refer to this "leveling up" process like beating a new level in a video game. With each new level or phase of growth in a startup, you'll need to use the skills you've already mastered plus an added new challenge. By the time you're a 100-person, Series C company, the number of questions asked (and the sheer number of people deciding things) may have increased by a factor of 10.

Admittedly, the role of an executive team also becomes infinitely more complex along the way.

It's likely that there are people on the executive team who are going through the process of company-building for the first time, and making decisions might seem scary. Part of why it can be helpful to have a board of entrepreneurs, VCs, and advisors is because these people can offer insights, observations, and pattern recognition that you may never be able to see from your vantage point. And certainly, by taking board seats on many of the deals we see at USV, we strive to stick close to our entrepreneurs throughout each step of this process to do just that.

But there are many great people at startups who will never have the privilege to be a part of board meetings or interact with investment partners at any VC firm. Those people still make decisions, too. Maybe not the kind that are discussed the board level, but still decisions nonetheless. These might include things like:

  • How should we design behavioral interview questions for our candidate process?
  • Should we migrate our system architecture to microservices?
  • What PR firm should we work with to announce our Series C?
  • Should the engineers sit next to the sales team, or on a different floor?
  • Would a rebrand make it easier for us to launch internationally?
  • Would this strategic partnership help us meet our OKRs this quarter?

No matter where you sit inside of an organization, you likely have some decision-making power.

Maybe you manage the office pantry and decide what type of food and snacks will give your colleagues the most energy to go about their days. Maybe you run IT and help prevent cyber-attacks or data breaches by educating your colleagues about safety best practices. Maybe you run your events marketing strategy and research the best conferences where you might set up a booth on the trade show floor to generate the highest quality leads for the sales team.

The single weight of any one decision may be quite small, (See: "Should we have both breakfast and lunch snacks in our fridge?") but the outcome of each choice matters. Over time, each "good call" contributes to an additive effect of a stronger, better company. And this is how we've designed the peer-to-peer effects of learning across the USV portfolio network. We want to empower everyone to make the best decisions possible by surfacing as many different perspectives as we can.

Rather than expect 1 person at the top to make the “right” call 100 times in a row, we've designed the USV Network around a second possibility: Can we help 100 people each make one slightly better decision than before?

In aggregate, by encouraging cross-pollination of ideas and experiences through all of the talented people across our portfolio, we believe we can leverage our network to serve as a resource to help all employees learn and grow.

In other words, when looking across our entire startup portfolio, we might ask: Can we help 10,000 people across 75 different companies to each make one better decision?

Each year, there are more than 150 touchpoints for people across this network to come together and share "lessons learned" at various in-person and remote events. Outside experts join us from time to time to time, bringing outside perspectives about management, leadership, or other common questions that arise. And dozens of people have stepped up to lead their own programming or even share back advice in one-on-one interactions with peers. In the end, we hope that by regularly convening people with similar functional roles across different companies, we might introduce new possibilities, solutions, and ideas -- one person at a time.

So, can the USV Network help make 10,000 better decisions? It'll be impossible to know for sure. But we're certainly going to try.

Come for an Action, Stay for the Community

If the last decade has been a time of digital abundance--more followers, more “friends,” more apps--2019 looks to be the year of paring down. 2018 saw the first significant decline in time spent on site for Facebook. Users are deleting the app and replacing it with tools to track screen time. But our desire to find community and connection, and to leverage technology to do so in more effective and efficient ways, hasn’t changed. As a result, the coming years will likely be a time of renewed opportunity in new forms of social systems, the kind that has been difficult to come by during the major platforms’ ascension and dominance.

This is not to say the incumbent platforms are going anywhere anytime soon. But there is likely to be a splintering significant enough to create new startup entrants with the chance to own consumer needs.

This splintering seems to be happening for two central reasons:

  1. Privacy and security. While events like Cambridge Analytica have drawn increased awareness to both the massive sets of data platforms like Facebook hold and the relative ease with which they can land in the wrong hands (let alone what the platforms can and likely will do with it themselves), privacy is still probably the lesser of the two core reasons for the degrading engagement. Increased privacy breaches that hit close to home for users will likely further point out that these platforms have created scale but not deep user trust--and emphasize the need to builded trusted brands to create true longevity. But, for now, its impact is probably lesser than that of reason #2.
  2. Users are no longer getting what they want (or as much of it.) The dominant platforms began with promises of curated communities, but have left us with crafted personas of what we want the entire world to see and a loose web of paper-thin connections. In the case of Facebook, this meant starting first with schools, then geographies, with real identities, initially verified with university addresses. But with an advertising-based model predicated on scale, success meant breadth so their purpose evolved to connecting everyone on the planet. They’ve accomplished this mission with billions of active users. However, the mandate of scale has made that initial premise of community on platforms harder to achieve. A decade plus in, this kind of broad connectivity is starting to seem more addictive but less fulfilling than many originally thought. Users are uber-in the know but feel emotionally disconnected. Lasting consumer businesses have to fulfill some set of human drivers and desires that, despite rapid innovation & technological change, basically remain the same--belonging, connection, curiosity. When they aren't being fulfilled, the products we use change, not the desires.

So what now?

The next wave of social systems will likely emphasize breadth and depth differently than the last--and may not look much like social networks at all. With the current networks, the horizontal nature of the platform is the product. Users come for the community. But it may be that with the next, the community is what keeps users long term engaged but is formed around another intent. Come for an action, stay for the community.

These platforms are likely to have four core things in common:

  1. Users share a common interest or objective (beyond connectivity.) Gaming is the vertical furthest ahead on these kind of new social systems and platforms like Twitch and Discord, in addition to certain games themselves like Fortnite, serve as great examples. Users come to watch a game, but stay to keep debating and discussing with a community of other gamers. Other verticals are earlier but show the same behavior. ShopShops* users sign on to watch livestreams of hosts shopping in retail stores and markets around the world, and to buy in real time. But, while they do, they wind up messaging with each other--responding to questions, sharing thoughts on the items, even splitting purchases (once a shopper only wanted one earring in a pair and another on the livestream happily took the other.) They come back as much to re-engage with the community as they do to find what they want to buy next.
  2. Users have skin in the game--they’ve in some way transacted or are on path to transact. This definition of transaction is broad--they purchase something, subscribe to something, or actionably engage with the brand before entering this group. This creates a deeper level of buy-in and, in effect, creates a “closed” community. There’s a lot of this behavior currently going on across platforms with the community building on Facebook and the transacting happening elsewhere. Duolingo’s* users connect in Facebook groups organized by the specific language they are learning to get extra practice, ask for help around tricky language rules, and share jokes. Embark, which sells an at home DNA test for dogs, has thousands of customers aggregating on Facebook to post photos and guess the outcomes of pending results. Owners with similar breeds connect with excitement, sharing tips. Increasingly, this community piece will be built into the platforms themselves rather than existing externally.
  3. Users in the best communities go through a transition from the specific to the broad. While users may join the network with a particular question or straight from a purchase, the best ones will expand the scope of the applicable quickly once the common ground has been developed. Dia & Co’s* most passionate customers, for example, aggregate in a group where they share best finds from their wardrobes but also ask for career advice and even plan vacations with women they only know through a mutual love of Dia. Glossier has the Into the Gloss community where passionate beauty enthusiasts come--first to talk about their favorite products (Glossier and otherwise), but then to go back and forth on travel tips and best self care post pregnancy.
  4. Multimodal business models--the best of these new networks will, over time, develop multimodal models that both take into account aggregated audiences as well as the services and products uniting them. Connie Chan at a16z wrote a great post on what we can learn from China about what some of these might look like. For the best ones, a multifaceted model that aligns business and customer combined with a platform that promotes transparency and depth of interaction will create a new level of defensiblity and stickiness.

If you are working on a business involving this new evolution of community, I’d love to learn more.

*ShopShops, Duolingo & Dia & Co are all USV portfolio companies

Nice To Meet You

Back in October I read a blog post penned by my now colleague Bethany Crystal about the launch of the USV Talent Network. At the time, I was consulting the French government on how they could attract more American growth stage talent to work at the most promising French startups. I sent a cold email over USV’s way explaining what I was up to and asking if they’d like to chat about talent, networks, and so on. Bethany responded with an enthusiastic yes and a job description of a new role she was hiring for.

Now here I am, just a few months later, penning my own blog post and introducing myself to you as the new Network Programs Manager at USV. So hey there! My name is Matt Cynamon (pronounced like the spice). I’ve spent my entire career in and around tech startups in the United States and abroad. I’ve launched two companies, and joined another that I helped grow from a handful of employees to a $400 million+ exit. Outside of work, I teach a class on how to get a job at a startup, I make mediocre, amateur films, and I spend as much time as possible in Prospect Park.

As the new Network Program Manager I’ll be launching some initiatives that unlock the value of the USV extended network (talent, alumni, partners, etc) to our portfolio of 75+ companies. I’ve been a massive admirer of USV since I moved to New York in 2009 and feel incredibly humbled to be doing my small part in bringing this network a little closer together. So, if you’re interested in working within our portfolio then check out all the open listings here and if you have ideas for how to improve the USV network don’t be afraid to reach out - [email protected] As is a testament to my being here, you never know where a friendly email might lead. You can find me on twitter @mattcynamon.

P.S. If you’re curious where our heads are at, here’s a list of books I’ve been reading to prepare for the work ahead.

The opportunity for learning & development at scale

Millennial employees are less drawn to the perks of their parents’ generation, like company-funded pensions, and more motivated by flexibility, autonomy, and access to the newest technology (cool offices help, too). One thing that has remained unchanged, though, is the desire to grow and learn new skills. Large, established corporations have invested in training for their employees for decades, offering dedicated in-house coaches to make sure their less experienced colleagues have the skills to conduct a meeting, manage personal and professional life, respond to a manager, and speak in public. Amazon offers an intensive, month-long training and leadership program prior to hire and AT&T opened AT&T University where employees can earn technical credentials like web and mobile development, data analytics and entrepreneurship.

This kind of training is desired by millennials, 82% of whom say that formal training from their employers on the job is important in helping them perform their best. However, many startups and smaller companies fail to provide it. A lack of development resources is not only to the detriment of the employee, since training is good for employers too; by improving retention (one study found that 94% of employees would stay longer at a company if it invested more in their career) and engagement (Gallup reported that 67% of employees were not engaged in 2016), companies can also bolster financial performance.

There is a clear opportunity to serve this need in the rising workforce at scale through platform and marketplace approaches that combine coaching and training with technology tools, thus broadening access to professional and personal wellbeing.

Demand in the tech industry started with CEOs searching for personalized coaching and has paved the way for coaches like Jerry Colonna and the Reboot team to be oversubscribed for months at a time. Now, increasingly, interest in coaching has extended beyond executives. Coaching is a large and fast-growing $10b piece of the overall learning and development (“L&D”) market. The question is: how can companies drive down cost enough to broaden access to the type of coaching and development training that was once only reasonable to spend on executives?

Technology may be able to provide leverage in three ways. First, automating logistical tasks such as scheduling allows coaches to work with more employees, driving up their per hour value and increasing their overall capacity. Second, finding an employee the right match increases her probability of being satisfied with the interaction. Third, gathering data and reporting allows for individually optimized training plans. Lin-Jin and Andrew Chen highlight how complex building curated (or semi-curated) marketplaces like coaching platforms can be, and the power of these networks if you succeed.

There are several L&D companies taking this approach. Some examples: BetterUp focuses on psycho- and behavioral analysis to help their customers achieve greater satisfaction at work with coaches. LiveCoach provides life coaches who help to provide structure and perspective in and out of the workplace. Ginger does too, over chat. Torch and GoCoach promote efficiency, engagement, and learning. NextPlay matches employees with each other as mentors. Hone focuses on delivering coach-led leadership development programs with a focus on distributed teams.

Another question is how much technology leverage can exist? Can technologies become the coaches themselves? On the one hand, you may want to get advice from another person (at least some of the time) because common to humans is an emotional inner life that allows for empathy and creates connection. On the other hand, if you’re interacting with a machine (and know that you are) you may be more open and honest about issues over which other humans might be judgmental. We’re still thinking through what aspects of the coaching itself can be impactfully outsourced to software.

A final question is whether it makes more sense for coaching and L&D platforms to go direct to consumer or to sell to enterprises. I tend to think that the latter makes more sense. With the war-on-talent in full swing, it’s possible that training and professional development will become a significant differentiator and selling point for employers when competing with the likes of Amazon and Google (Google offers personal coaching to every new hire in its competitive APM program).

If you’re working on ideas in this space we’d love to hear from you, please reach out.

Our First Network-Wide Diversity & Inclusion Survey

Earlier this fall, Lauren Young on our Network Team collaborated with companies across our portfolio network on our first-ever USV Network diversity and inclusion survey.

One of the things we heard through our monthly USV Network diversity & inclusions discussions is that, while baseline metrics exist at larger organizations like Google, Facebook, or even Etsy, it’s hard to get a sense of where a 100-person startup might set their baseline.

Our goal was to understand a few things:

  • How are startups doing in terms of promoting diverse and inclusive work environments?
  • What initiatives have been the most successful at rolling out a D&I strategy?
  • Are there any patterns or trends we can expose within our network to share any best practices?
  • What, if anything, can we do from our place as a VC firm to help?

In the end, 38 companies* (just over 50% of our active portfolio) participated in this survey, with 55% of respondents falling between 31-100 employees.

While this is obviously still a small sample size of startups, we wanted to share a few patterns that we noticed as well as share the survey questions that we asked among our network, in case you’d like to repeat this with your own portfolio or organization.



What we learned:


Companies over 100 employees are most likely to have diverse executive teams.
Of the 38 participating companies, 61% reported having women and 57% reporting having people of color* in the C-suite. While we didn’t ask about any further breakdowns (for instance, whether there is equal representation among men and women among these exec teams), we did find that companies with more than 100 employees were the most likely to have women and people of color in the C-Suite than smaller organizations. We did not survey on under-represented minorities, which we will fix in future surveys.

There's an opportunity at the director level.
In our survey, we found that only 16% of companies have at least 50% representation of women at director level roles and above. Additionally, 6 in 10 companies have less than 20% representation of people of color* in Director level roles or above

Creating inclusive policies seems to be the most common first step.
When we asked companies what steps they are taking to promote diversity and inclusion initiatives internally, we found that writing an anti-harassment policy, creating a code of conduct, and tracking diversity metrics were the most common practices. 71% of companies have at least started implementing anti-harassment policies and a code of conduct. Some of the least explored initiatives included creating an inclusion communication guide and building company-wide “affinity groups.”

The most impactful changes come from deliberate, intentional effort.
We asked our companies to share their biggest “lesson learned” from the efforts they have undertaken thus far. The most common theme that emerged was that it takes repetition, intention, and strong communication to get this right

It's hard to find the time to make it a priority.
One of the free responses questions in our survey asked about the biggest challenges our companies faced in order to get diversity right. Some common answers we heard included building diverse candidate pipelines, integrating D&I into the organizational DNA, committing to spending time on D&I programs.

There's a lot we can do as VCs to support this work.
When we circulated these results internally at USV, we were most eager to explore how we can foster increased opportunities to amplify these efforts among our portfolio. A few great ideas emerged from this research for ways that our firm can help with recruiting efforts, building community, and offering cross-company training opportunities for our network. We intend to explore some of these ideas in 2019 and beyond.


What’s next for us

While we recognize that this survey is only an early step for us in better understanding the D&I needs for companies across our network, we hope that our companies and other communities can learn and build upon this work, too. To that effect, we’ve open sourced the survey questions that we asked to our network, including a few comments of places where we think we can improve in the future.

We welcome the opportunity to continue collaborating with internal and external partners on this work.



Special thanks to Kapor Capital, for collaborating on the question set for our survey, and to Jasmine Shells and the whole team at Five to Nine, an employee engagement solution that uses analytics to improve company culture, who helped us analyze these results.

*Our data collection process:

  • We received 38 total responses to this survey, representing just over 50% of our active portfolio at the time it was conducted (October 2018). The demographic breakdown of company sizes reflected in the responses is comparable to the overall breakdown within our active portfolio today.
  • We defined “people of color” as a person of non-white or European descent and did not survey about underrepresented minorities, which was an oversight. In follow-up surveys, we intend to better understand how both populations are represented across the network.

Therapeutic Computing

As part of the broadening access to wellbeing thesis, we have been searching for projects that bring down the cost of therapy and coaching. 8% of adults and 12% of teens (and as high as 20% of teen women) suffer from depression in the US, but traditional therapy costs $75-150 an hour and online therapy like BetterHelp and Talkspace cost $150-$300/month, which are unaffordable for most US households.

One way to approach this challenge is building AI chatbots that act as personal therapists. This is a great approach, but I also think there is an opportunity to build a new kind of therapeutic relationship between a human and a computer that looks different than the traditional relationship between a human patient and a human therapist. Such a model would lean on what a computer can do really well that would be difficult for a human to do, such as ingesting a large amount of unstructured data and deriving insights.

One such model that seems potentially very interesting is software that passively ingests messaging, location, activity, sleep and phone usage and offers observations and guidance in a helpful (and privacy preserving) manner.

One example could be a calendar that passively notices that you haven’t been attending your scheduled plans and helpfully asks if you’d like to set aside time to journal or call a friend. Alternatively, a calendar that has access to your messages and offers you suggestions for what to talk about with a friend before you go to meet them.

Another example could be a keyboard that offers inline observations about your mood and reminds you to stop and breathe if you're about to send an angry message.

A third example could be a fitness app that notices that you haven’t left your house for a few days and offers you a nearby activity you might like to walk over to.

People are building all sorts of passive observational guidance tools that could potentially be very impactful:

  1. Mei and Actual are both building messaging clients that offer advice for how to better connect with the people you talk to. Mei, for example, after watching my messaging threads, suggested I take more notice when my friends get worked up about something.
  2. Maslo and Scribe are both journaling apps that also over time offer insights about your mood and surface the common themes of your journal entries.
  3. Sonic Sleep Coach is an alarm clock that suggests personalized improvements for how you can get better sleep.
  4. Just Not Sorry is a Gmail plugin that helps you draft emails that more strongly communicate your thoughts and ideas.

I’m sure there are others.

Two benefits of a software-only solution are that software is stigma and judgement free, and can be low cost for the user. However, it’s important to be careful that these observations and suggestions are helpful and not hurtful, and so there could be systems that start with or even forever have humans in the loop.

One big question is how to build therapeutic, personal software that is helpful without being creepy. The answer may be to set up technological or business structures that guarantee the user’s trust. In a recent USV meeting, we talked about this idea by Muneeb Ali as the difference between a business with the mission Don’t Be Evil and a business with a structure that Can’t Be Evil.

One example of a Can’t Be Evil technical structure is running data analysis and machine learning on device so that the company has no access to users’ private information. Another is encrypting data on device a la iMessage so that even if data is stored on the company’s servers, the company does not have access to it.

Another example of a Can’t Be Evil business structure is building a governance system that gives some of the control over to users. Another is having a business model that aligns the users’ interests with the business’s such as users paying directly for the product via a monthly subscription, even if it is low cost, such as $1/year.

I’m curious to hear others’ thoughts on therapeutic products that work through passive observation and interested to see what others are building. It seems important to build a world where anyone, anywhere can get access to tools for wellbeing and finding new models to deliver those tools feels like a big piece of solving that access puzzle.

Looking For Syllabus 2.0

Becoming a fast expert in a new topic is still a big challenge.

It’s a big unmet need: there are 26 million students taking courses on Coursera, but only single digit percentages of them are finishing the courses they start. People want to ramp up to proficiency on new topics, but are missing a compelling way to do that quickly.

There are a lot of open resources available on the web, but finding, qualifying, and navigating them--let alone combining them into a self directed system--is a challenge. That's because the tools we use today to search the web are designed for quickly finding facts, not for guiding you on a learning journey. For self-learning new topics, we need something that's more like a coach than an encyclopedia.

In the classroom, the syllabus plays that role. The syllabus is a learning map. It tells you what you should read, in what order, and what is the broader theme you should be thinking about at each step. It also tells you how many weeks the whole thing will take. (PS if you're into old school syllabi, check out The Open Syllabus Project at Columbia University).

There seems to be a big opportunity to reinvent the syllabus and create best of class learning guides crowdsourced from the already existing open materials on the web.

There have been several attempts already to curate online resources for learning new topics. Usually they take the form of a list of links. The problem with the list of links approach is that they are static and they are inefficient. You don't need to read a whole link to get the main point, you want to curate little bits and pieces of open resources: 30 seconds of this podcast, a minute and a half from this youtube video, just these 4 paragraphs from this article.

The thing that is closest to a modern internet syllabi is Susan Fowler's guide for learning physics (it's really amazing, go check it out). What if you could have that type of curated guide for many topics that gets updated by the community over time, with inline discussion with other learners?

I think Syllabus 2.0 could look something like this:

And I think this model could work for a variety of topics, from How to Appreciate Baseball, How to Become An Expert In Machine Learning Without Doing Math, and How To Think About Cryptonetworks. We've created a sample syllabus for this last topic so you can see what we envision in action. It curates 8 hours of podcasts, talks and blog posts into a 30 minute guide. There are inline comments so that learners can have discussions and ask each other questions, and it’s on GitHub so that anyone can suggest changes.  

It may be useful to break out topics into a 5 minute guide, a 30 minute guide and a 3 hour guide. And because learners approach topics with different levels of background knowledge, maybe at the top of each syllabus there is a quick 5 question placement quiz that determines where in the syllabus a learner should start. Someone that can answer some basic questions about a topic can start in the middle of the syllabus while someone who is learning about a topic for the first time should start at the beginning. The syllabus should also adapt to each individual user’s style of learning and the user’s intent –– whether they are learning for fun or to achieve a career-specific goal. And the syllabus could also incorporate quiz questions throughout to close the loop and double check the learner’s understanding.

How would a syllabus project turn a profit? One possibility is that the syllabi are the free content at the top of the funnel and as people are more and more serious about learning a topic, they can pay to join an online class, fly out to take a week-long certificate-granting seminar or get matched with a learning coach. People should be able to learn for the sake of learning for free, and if their goal is to learn in order to change their career or level up professionally, they could have the option to pay to expedite and certify their learning.

There are a few learning coach models that I really like beyond traditional 1-1 coaching. There was a project in 2007 by Sean Dockray called The Public School where if enough people were learning a certain topic, they could come together and find a tutor that would work with all of them in a self-organized class. Another model I like is what Ray Batra is doing with learning gyms. He is building co-learning spaces (think WeWork, but everyone there is actively learning something) where there are coaches there to help. Another model that is less like coaching but still effective is if after reading a syllabus that interested you, you could sign up for a weeklong in-person intensive course on that subject led by an expert.

There are five projects I know of doing something in this space: Learn Anything, which lets people upvote the best resources for learning any topic, Gooru, which lets teachers use existing content on the web to create courses for K-12 content, Holloway and Golden, which are creating open source guides, and Hyperreadings, which is a permanent archival library plus curated reading guides on top. I am sure there are more.

We're actively looking for the learning guide of the future. A large part of our newest investing thesis is about trusted brands that broaden access to knowledge, and an exciting path to achieving that goal may be to modernize the syllabus and bring it online. If you are working on a better syllabus, reach out: I'm [email protected] and I can't wait to learn from you.

“The most valuable thing humans have ever created is our knowledge.” – Juan Benet


Welcoming Dia&Co to USV

Today, Dia&Co announced their Series C round, which we recently led out of our Opportunity Fund. The crux of USV’s Thesis 3.0 is about backing trusted brands that broaden access, and this is exactly what Dia is doing.

Dia is a commerce and community platform for plus size women. Their first product has been a curated box with a try-at-home model, where customers keep what they want, send back the rest, and receive increasingly personalized product over time as Dia learns their preferences and styles. But Dia is based on a customer and not a model--how they serve her will evolve; who they serve will not.

There are two core reasons that make Dia particularly exciting as an addition to the USV portfolio. First, the size and current state of the market they are focused on. Second, how well positioned this team is to tackle this problem and the ability they’ve shown to create not just a transactional business but a large, fast scaling, emotionally engaged community.

While the plus size category has been largely ignored by retailers, it has evolved into an ideal opportunity for new brands to emerge that allow high intent, passionate customers to transact in ways they are asking for. It is rare and getting rarer to find real market gaps in the commerce world. Usually, there’s opportunity to improve on the product quality or composition, user experience, or price point. But in plus size there’s a complete supply and demand imbalance.

The stats tell this market’s story cleary. Plus size is defined by size 14 or up--a category that encompasses at least 68% of American women. But a look at 25 of the country’s biggest retailers showed that only 2.3% of their assortment was plus size. It makes up about 17% of retail overall, generally tucked away in basement levels of department stores, unmoved or reimagined for decades. (Nadia, CEO of Dia, and the Gotham Gal go into some of this during this podcast, which is excellent.) Today, even with poor experience and lack of supply, that’s a big market--about $21.5B in 2016 and growing faster annually than any other segment in retail. The potential however is meaningfully bigger.

Better yet, this customer isn’t quiet--she’s high signal. She likes and posts and comments about what she’d want to buy if it was offered. Just like many straight size women, she’s passionate about fashion and trends, even if she’s rarely able to reflect that with her dollars. Nadia Boujarwah, CEO of Dia&Co, knew this well because she’d been this customer her whole life. So when she set out with her co-founder, Lydia Gilbert, to build Dia, she understood the rarity of the opportunity first-hand--a massive market with high demand, high intent, and extremely low supply.

When I first met Nadia and Lydia 3.5 years ago, Dia was in very early days. Nadia and Lydia started by buying the clothes and shipping them to customers they emailed with. It was low tech, high touch, and not yet scalable. But the deep customer desire was overwhelmingly evident. Women were asking when their next box could arrive before their previous one had even made its way back. The more boxes they got, the more they bought--Dia could gain their trust and convince them that in a retail world where they'd always been forgotten, here they were heard and celebrated.  

Several years later, Dia has seen rapid growth and has built scalable infrastructure. Deep data science and algorithmic recommendations complement the army of passionate customers turned stylists. An easy to use interface has replaced email. And sophisticated inventory management, private label creation, and complex reverse logistics make the backbone of best in class merchandising and operations. But, most of all, this trust in the brand has only strengthened and remains the highest potential element of what’s now a big business. For many of their customers, Dia has become more than a commerce platform, but a much needed tribe. Recently, a group of Dia women gathered in Nashville for vacation. They had never met in person but had long been communicating on the Dia&Co Facebook group. In each other they had found commonality and friendship; in Dia, they found a community they feel a part of, and products that they not only wore but felt proud in. Whenever they receive a box in the mail, they immediately post each item to the group to solicit opinions. They are honest with each other on what to keep and what to ditch, but always supportive.

That belonging isn’t unique to the women on the trip. Its seen throughout Dia's base of customers and gives them license to increasingly deepen mind and wallet share. That ability is at the heart of the best consumer transactional businesses and the signal of a real trusted brand.

We are thrilled to welcome Nadia, Lydia, and the Dia&Co team to the USV Network.


Differentiated user experiences are a cornerstone of the defensible businesses that USV has invested in for over 10 years. In the most recent articulation of our thesis - 3.0 - we also stressed experiences and trusted brands that expand access to, for example, knowledge. Trusted services are those that have business models that better align the interests of businesses and their customers.

What would such an experience and aligned business model look like in the world of online publishing, that to date feels like it has replicated the offline model in a way that is cumbersome and decreases consumer value through clunky and increasingly invasive experiences? One that was resolutely consumer first by being clean, fast, and safe. One that wouldn't require users to change how they access content in order to benefit from the experience. And one that did not use a paywall.

Scroll is a consumer and publisher service that satisfies these criteria by providing frictionless access to content by removing all ads and ancillary content, making pages clean and fast. It is no surprise that a specific user experience was the focal entry point for this company - the founding team’s background includes Chartbeat, Foursquare, Spotify and Google (some of those companies we have previously founded and backed). USV is excited to be leading Scroll’s Series A financing, announced today, and joined by Samsung Next, Bertelsmann, Gannett, Axel Springer, The New York Times and Uncork Capital.

Scroll is a vastly better way to read or watch, across multiple channels - an open platform that puts consumers in control by offering bundled access (for $4.99 a month) to ad-free and mobile-friendly experiences on content sites around the world. When you visit a Scroll partner, that site automatically delivers an experience with no interruptive ads, pre-rolls, or links to baldness cures at the bottom of articles. Users control their own data through an API, and they can decide what they want to do with it - from requesting their data be anonymized to sharing their history with other services. Using mobile apps like Twitter, Reddit, LinkedIn or Facebook or just reading in Chrome, Scroll seamlessly syncs content across devices and is format-agnostic. One can start reading an article on a laptop, continue on a phone on the subway, and then finish as an audio story in a car.

Dozens of publishers are currently part of the network, including Buzzfeed, Vox Media, MSNBC, Business Insider, The Atlantic, Slate, Fusion Media Group, The Daily Beast and more. We believe this model empowers, not disintermediates, publishers by giving them an important choice in addition to their ad-supported business. As important, publishers using Scroll will make more money than using an ad-only model.

Now, more than ever, we need an Internet that feels built for us and a free press able to thrive. Scroll firmly believes in this vision, and USV does too.

Introducing the USV Talent Network

Today we’re excited to announce the launch of our new USV Talent Network.

Our USV Network is a startup community of more than 9,000 employees across 75 companies all over the world. Each company leverages network effects to build stronger businesses and broaden access in differentiated ways.

We operate our portfolio network in a similar way and support every employee in finding access to the people, information, and resources they need to do their jobs better. Because this collective knowledge base increases as our portfolio grows in size, working for a USV company broadens your network each time we make a new investment.

This initiative is the first time we are deliberately seeking to expand the reach of our USV community beyond our current portfolio. We hope to extend our collective footprint even further -- to seek out candidates our companies may want to hire, to identify best-in-class leaders who would like to be a part of our community, and to invite alumni (people who used to work at USV network companies) back into the fold.

Built in partnership with, once you’re accepted to this network, you can request introductions to companies across the USV Network when you see jobs that interest you. Even if there aren’t any current job openings that fit your criteria, you’ll be able to signal your preferences and receive notifications when new roles open up in the future.

If you are interested in working for one of our portfolio companies (either now or sometime in the future), we invite you to apply to join our extended talent community. Over time, we’ll surface interesting jobs to you and invite you to engage with our current portfolio by sharing your expertise or joining us at events. Part of the benefit of being in the USV Talent Network is the ability to connect with the companies directly instead of having to compete with every other person applying to a job.

As an employee of one of our portfolio companies, you’re invited to participate in cross-company peer groups and learning opportunities that support you through all phases of your personal growth at a startup. These include:

  • Meet each other: Get to know peers in your domain and familiarize yourself with common challenges and approaches that others have taken in your role
  • Level up: Participate in cross-company manager training or leadership development tracks to increase your sophistication of startup know-how and establish a peer group
  • Give back: Share case studies and advice from your experiences with our network as an advisory role or peer cohort leader of more junior network members

Throughout the year, the USV Network facilitates more than 150 events (both in-person and digital) to bring together peers in similar domains at different companies and encourage cross-company learning opportunities.

It can be scary to join an early-stage startup. There’s a lot of inherent risk through the discovery process of identifying product-market fit and dozens of complications along the road to scaling a business. If you’re the joining a company sub-50 employees, it’s likely that you’re the only one at that organization in your functional domain area. That feeling can be pretty lonely.

When you get a job in the USV Network, you’re not just joining a company; you’re joining a community. If you’d like to be a part of this community now or sometime in the future, we invite you to join us today and apply here.

Business Model Innovation in Healthcare

Recent conversations around healthcare have surfaced words like “crisis” and “collapse”. Besides being confused by the existing system, 20% of Americans can no longer afford basic healthcare (including their insurance premiums), leading to a higher number of people who are uninsured or underinsured. Now, unbundling is enabling a wave of new business models as the healthcare system shifts towards more virtual, independent, local points of care (an idea introduced to us by Dave Chase in his book, “The Opioid Crisis Wakeup Call”).

These new approaches generally fall into one of two buckets: change of infrastructure (who pays, when they pay, how they pay) and change of venue (from in person to online). The emerging business models tend to be much simpler and instead of operating in bundled networks, there is an increase in independent clinics and care delivered through mobile phones.

Change of Infrastructure

Four novel models of care are emerging: direct to consumer, direct to employer, value-based, and group cost-sharing for catastrophic events. Recently, we’ve been looking at this first category, however, they could all broaden access with the advent of unbundled health services.

Direct to Consumer Healthcare

Direct primary care practices are opening up independently of a hospital network or insurance company. In this model, each physician operates as a separate entity and can decide how to charge their patients. Many, such as Unorthodoc and Gold Direct Care, seem to be using a subscription model and some practices use different tiers of pricing depending on the level of care that a person needs (usually around $30 to $120 per month). Another example, SparkMD, offers custom packages including family rates, house visits, and telemedicine.

While well-loved concierge practice One Medical prices their membership package at a premium (~$150 per year on top of the cost of care), DPC clinics generally hustle, trying to provide affordable options wherever possible. They broaden access. The subscription model for primary care (Healthcare as a Service, or HaaS), whether in person or virtually, aligns incentives more naturally between the patient and provider.

If each primary care physician - which many view as the center of healthcare - operates an independent practice and is able to communicate with local specialists to find the best price for their patients, then smaller microeconomics could start to form in each city or town that allows patients to make decisions with full price transparency. In today's more centralized model, physicians that are part of larger network tend to refer patients to facilities in-house even if they are more expensive.

Direct to Employer Healthcare

Instead of decentralizing healthcare to local neighborhoods, the system can be broken down to a different unit: employers. Today, employer-based healthcare tends to prioritize convenience over affordability. However, with more unbundled services and specialists, what if every employer had its own system for patient care and external referrals? Euphora, Paladina, and Apostrophe are examples of companies enabling this model. In the case of the latter, in particular, by replacing third party administrators for self-insured employers and saving employers money in the process. Elation proposes that employees are more likely to seek out preventative healthcare - and thereby lower downstream costs - through an on-site clinic.

Employers are one of the most incentivized players to keep healthcare costs low and employees healthy. If primary care operated in-house, employers would manage the workflow and payments through a platform like Eden, for example. More interestingly, this can be extended beyond unbundled primary care. Sano Surgery offers self-insured employers access to medical professionals in 13 specialty verticals and even allows them to pre-purchase medical services, which fundamentally changes payment dynamics.

Employer-based healthcare doesn’t necessarily solve the ‘access to healthcare’ problem because employer-based insurance covers only 56% of the US population today. This begs the question, should companies (or even, can companies), in addition to their mission, be focused on healthcare as well?

Value-based care

In value-based care, payment is based on health outcomes for entire episodes of care rather than solely on initial treatment plans. This is straightforward for commoditized parts of healthcare, such as a routine surgery, where a la carte pricing works. A great example of a simple business model is the Surgery Center of Oklahoma, where patients know prices and pay up front for each “episode”. The SCO website lists medical surgeries as food would be listed on a menu. As patients continue to visit from in and out of state, their goal is that more centers and practices will adopt this practice, making the back-and-forth with insurance companies obsolete.

Thinking about long-term illnesses - for example, diabetes management, cancer, or asthma - becomes more complicated. For an independent specialty clinic like Pure Cardiology, value-based care might require making a plan and having the patient pay when specific milestones are reached. While this solution could result in delayed payments for doctors, they can ensure that incentives are aligned for both short-term and long-term episodes of care.

New Forms of Insurance

Medical cost-sharing groups, where people pay to support others in their community, are appearing more often. Medishare, Sedera and Liberty HealthShare are examples of cost-sharing within local communities, neighbors if you will, that combine a non-medical community - either geographic or religious - with their medical insurance for both routine health issues and catastrophic events. In some instances, these groups may appear closer to non-profits than large insurance companies: Liberty HealthShare’s decision guide emphasizes that they are not “insurance”, and are incentivized to work towards decreasing the healthcare burden on the whole group. Still, for-profit innovation is occurring on the insurance side.

Change of Venue

Virtual Primary Care

Many companies are working towards a future with at-home testing, processing prescriptions with one tap on your phone, and virtual appointments (including USV portfolio companies Modern Fertility and Nurx). More fundamentally, there could be a subscription option for primary care for those that only interact with their doctors by mobile phone unless otherwise necessary. Companies like Sherpaa and Nurx are changing the venues of care to text-based conversations, at-home testing, and remote monitoring. Asynchronous, structured data collection thrives in a decentralized model of healthcare because it is cheaper and time efficient, allowing each doctor to help more patients and decrease the points of friction in getting care.

A pure VPC model eliminates fixed costs associated with brick and mortar expansion and is able to focus resources on reaching more patients, recruiting more doctors to their platform, and improving the experience for current patients. Payments on a subscription basis allow doctors to get paid more consistently rather than waiting for insurance companies to process claims and paying overhead costs to negotiate reimbursements with their billing offices.

Pop-up clinics

Mobile, pop-up clinics in different healthcare verticals (from urgent care to specialized medicine) are also growing. An open question regarding broadening access is whether this model is scalable to all parts of the country, or whether it perhaps favors population dense cities. Two examples in women’s health are Kind Body and AskTia. Women can go for a regular checkup or with specific questions related to fertility, IUDs, etc, and receive a highly specialized experience. This can feel much simpler from the patient perspective, as these models often rely on subscription payments or disclosed, fixed pricing for each service. Another example, specific to cardiology, is Heartbeat, where people get direct access to a fully-digitized boutique cardiology clinic. Here, the venue of care is shifted here from hospitals to pop-up clinics, that are smaller and easier to scale.

In Summary

Health-related issues are confusing as is, so the system around it should be as simple as possible. New business models are lowering the barriers to entry for users to get access to high-quality care. Cost is driven down through technology and healthcare is made simpler through new unbundled business models, streamlined payments, and better user experiences. As companies continue innovating in this space, both in terms of business model and venue, the future of healthcare is starting to become clearer. We’re excited for that future.

The Myth of The Infrastructure Phase

A common narrative in the Web 3.0 community is that we are in an infrastructure phase and the right thing to be working on right now is building out that infrastructure: better base chains, better interchain interoperability, better clients, wallets and browsers. The rationale is: first we need tools that make it easy to build and use apps that run on blockchains, and once we have those tools, then we can get started building those apps.

But when we talk to founders who are building infrastructure, we keep hearing that the biggest challenge for them is to get developers to build apps on top. Now if we are really in an infrastructure phase, why would that be?

Our hypothesis is that this is not actually how things play out. We are not in an infrastructure phase, but rather in another turn of the apps-infrastructure cycle. And in fact, the history of new technologies shows that apps beget infrastructure, not the other way around. It’s not that first we build all the infrastructure, and once we have the infrastructure we need, we begin to build apps. It’s exactly the opposite.

A big part of why this is even a topic of conversation is that everyone now knows that “platforms” are often the largest value opportunities (true for Facebook, Amazon/AWS, Twilio, etc.) -- so there is naturally a rush to build a major platform that captures value.  This may be even more true in the distributed web where value often -- but not always -- accrues in the protocol layer rather than the applications that sit on top.

But, as we will see: platforms evolve from an iterative cycle of apps=>infrastructure=>apps=>infrastructure and are rarely built in an outside vacuum.

First, apps inspire infrastructure. Then that infrastructure enables new apps.

What we see in the sequence of events of major platform shifts is that first there is a breakout app, and then that breakout app inspires a phase where we build infrastructure that makes it easier to build similar apps, and infrastructure that allows the broad consumer adoption of those apps. Kind of like this:

Apps and infrastructure evolve in responsive cycles, not distinct, separate phases.

For example, light bulbs (the app) were invented before there was an electric grid (the infrastructure). You don’t need the electric grid to have light bulbs. But to have the broad consumer adoption of light bulbs, you do need the electric grid, so the breakout app that is the light bulb came first in 1879, and then was followed by the electric grid starting 1882. (The USV team book club is now reading The Last Days Of Night about the invention of the light bulb).

Another example: Planes (the app) were invented before there were airports (the infrastructure). You don’t need airports to have planes. But to have the broad consumer adoption of planes, you do need airports, so the breakout app that is an airplane came first in 1903, and inspired a phase where people built airlines in 1919, airports in 1928 and air traffic control in 1930 only after there were planes.

Sometimes all the infrastructure you need is a beach and some spare parts.

The same pattern follows with the internet. We start with the first apps: messaging (1970) and email (1972), which then inspire infrastructure that makes it easier to have broad consumer adoption of messaging and email: Ethernet (1973), TCP/IP (1973), and Internet Service Providers (1974). Then there is the next wave of apps, which are web portals (Prodigy in 1990, AOL in 1991), and web portals inspire us to build infrastructure (search engines and web browsers in the early 1990’s). Then there is the next wave of apps, which are early sites like in 1994, which leads to a phase where we build infrastructure like programming languages (PHP in 1994, Javascript and Java in 1995) that make it easier to build websites. Then there is the next wave of more complicated apps like Napster (1999), Pandora (2000), Gmail (2004) and Facebook (2004) which leads to infrastructure that makes it easier to build more complex apps (NGINX and Ruby on Rails in 2004, AWS in 2006). And the cycle continues.

We also see this pattern with our most recent iteration of mobile apps: first we had a suite of popular mobile apps that relied heavily on streaming video: Snapchat (2011), Periscope (2014), Meerkat (2015), and Instagram Stories (2016). And now, we are seeing companies building infrastructure that makes it easy for mobile apps to add video: Ziggeo (2014), (2014), Mux (2017), Twilio Video API (2017), Cloudflare Stream (2018).

This cycle also correctly explains the sequence of events in Web 3.0. We start with the first breakout app: BTC (2008), on top of the Bitcoin network (as the first infrastructure), followed closely by Silk Road (2011) as the most infamous early crypto app. This inspires new infrastructure like Sidechains and Drivechain (2015), Ethereum Smart Contracts and ERC20 (2015), Lightning (2015) that make it easy to build new apps, and infrastructure like Coinbase (2012) and Metamask (2016) that enable consumer adoption of these new apps. This new infrastructure then enables the next wave of apps: tokens/ICOs (2017) and early dapps (Rouleth and vDice in 2016, CryptoKitties in 2017), which inspire new infrastructure: Infura (2016) and Web3js and Zeppelin (2017). We’re now waiting for the next big apps that will help guide the next wave of infrastructure.

The Adjacent Possible

The common theme in the development of each major platform (electricity, cars, planes, the web, mobile, etc.) is that we build what we can given the tools available to us at the moment.  In Where Do Good Ideas Come From, Steven Johnson refers to this as The Adjacent Possible.  In other words, you can open the door to the next room, but you can’t really skip steps and open the back door from the front porch.  It is hard to successfully build infrastructure that is too far ahead of the apps market.

Each time the apps => infrastructure cycle repeats, new apps are made possible because of the infrastructure that was built in the cycles before. For example, YouTube could be built in 2005 but not in 1995 because YouTube only makes sense after the deployment of infrastructure like broadband in the early 2000’s, which happened in the infrastructure phase following the first hit dot com sites like eBay, Amazon, AskJeeves and my favorite, Neopets.

Chris Dixon and Fred Wilson talk about this concept in a recent episode of the a16z podcast. Chris has a board game from the dot com era called Dot Bomb that makes fun of the silly dot coms of the late 1990’s. And what he points out is that all the ‘silly’ ideas of the dot com era are now the billion dollar unicorns. What is now possible several app => infrastructure cycles into the internet made no sense just one or two apps => infrastructure cycles in.

That is the crux of what we mean by the myth of the infrastructure phase -- if we think about an “infrastructure phase” divorced from the apps that will use it, we run the risk of building too far ahead, in a speculative vacuum.  We need the cycle of apps=>infrastructure=>apps=>infrastructure to keep us honest.

As there are more and more cycles in each new platform it gets cheaper to build and use those apps. Building in 1995 would have cost us many orders of magnitude more than it would cost to build today, and creating Web 3.0 apps costs more in cash, effort and time today than it will 15 years from now.

Development Frameworks Versus Investing Frameworks

Putting our investor hats on for a second, it’s important to distinguish between technological frameworks that explain when something can be built, and investment frameworks that explain when something can be a good investment.

The apps=>infrastructure=>apps=>infrastructure cycle explains when apps or infrastructure can be built, but doesn’t necessarily explain when to invest in apps versus when when to invest in infrastructure.

Take light bulbs for example. Yes, they were invented before the grid, but looking at it from an investor perspective, no one sold a lot of lightbulbs until the grid was in place.

Wrapping Up

One question we had is: why is it that apps come first in the cycle, and not infrastructure first? One reason is that it doesn't make sense to create infrastructure until there are apps asking you to solve their infrastructure problems. How do you know that the infrastructure you are building solves a real problem until you have app teams that you are solving for? It will be a challenge to build crypto infrastructure now until there is a breakout crypto app that other developers want to emulate and need better dev tools and infrastructure to do so.

There is a narrative in the crypto space that first we need to build great tools, and once we have the tools, then we can build apps. But what we hope to have shown is that in other platform shifts, we are able to build the first few apps before there are great tools (though it is more cash and time intensive), and then those early apps inspire us to build tools.  And the cycle repeats.

Happy building. 

Update (10/5/18):  This post has gotten a lot of attention, has generated some great discussion, and has produced some useful feedback. 

First: duly noted that we spent most of our time here looking backwards at historical precedent, and thus that our diagram on the decentralized web / web3 / crypto was a) admittedly thin, and b) really just focused on the ETH ecosystem.  We have updated that diagram to be a little more clear, and to include important concepts from the BTC ecosystem.  Thanks to Dennis Porteaux for the excellent analysis on this.

Second: our favorite piece of feedback is that crytpo networks, in fact, really blur the line between apps an infrastructure, due their open and interopable nature. That is one of our favorite aspects of them.  So, for instance, an "app" (like CryptoKitties, or any smart contract, or Bitcoin itself) can be infrastructure if someone builds on it.  Of course, there are components of these networks that are **only** infrastructure (Lightning, Zeppelin, etc), but the line is blurred.  Whereas in the past a platform (like Amazon or Facebook) had to make a conscious decision to open up APIs and become a platform, crypto apps are generally open and interoperable from day 1.  This only makes the apps=>infrastructure=>apps=>infrastructure cycle tighter.  Thanks to Denis Nazarov and Jutta Steiner for really articulating this.


Unbundling Healthcare

USV has been investing in health-related technologies for about five years. Our approach can be characterized by the notion that digital processes, made more accessible by the internet and mobile access, could over time transform the cost and delivery of medical care. In the last couple of months, our conversations around healthcare have surfaced some new investable themes, which I’ll discuss in a few posts over the next month. The first such theme is the idea that healthcare can be “unbundled” into distinct services, a pattern we have seen in other markets (where niche sub-markets develop for specific services). So, in the case of healthcare, what is there to unbundle from? For one, large hospital networks and insurance companies, where most of the control and pricing power currently reside.

The last decade has seen unbundling in the financial services industry that serves as a relevant analogy. Financial supermarkets became popular in the 80s when large banks started offering one-stop shopping for a complete package of financial services, anything from mortgages to credit cards. In this model, the sale becomes around convenience rather than value as customers buy a complete package, even though each service may not be the best in breed or meet the specific needs of an individual consumer. With this, came a lack of price transparency as each service’s individual price is obscured by the combined package. The connected Internet and later mobile services, have made it easier to manage an unbundled package as well as increase the quality and personalization of each service on its own. In finance, we’ve observed that in the broader market, as well as within our own portfolio of fintech investments, with Funding Circle (peer-to-peer lending platform for small business), Stash (personal investing), CircleUp (equity financing for consumer products), Stripe (payments), and KickStarter (crowd-funding). Technology made processes simpler so that individuals could find specific things they need easily and affordably from market-specific providers.

Perhaps healthcare is now seeing the beginning of unbundling, catalyzed by the same use of technology which allows companies to focus solely on user experience in specific verticals. Some of the first pieces of unbundled healthcare to get venture momentum were direct-to-consumer brands such as Hims, Roman, Keeps, and Nurx. These are examples of companies using technology - such as telemedicine, asynchronous chats, and structured data - to lower costs and reduce friction in various niches of healthcare via a simple subscription business model. For example, Nurx gives users control of their health at every step of the process and delivers birth control and Truvada for PrEP in a timely, cost-efficient manner. Along a similar vein, at-home testing companies such as 23andMe, Modern Fertility, and Scanwell are bringing lab-grade testing services to people’s homes at a lower price point by leveraging technology. For primary care, companies like Sherpaa are using virtual primary care, while Doctor on Demand and Teladoc started by unbundling to serve individuals through telemedicine. Now other companies such as Vera Whole Health, CareMore, and PeakMD are coming up to serve employers with their primary care needs through local care centers.

Then there are examples of unbundling in parts of primary and specialized medicine that have not received as much attention yet, perhaps because their interactions are primarily conducted offline. However, these companies are still a part of the trend towards unbundling from large hospital networks and functioning as individual nodes in the broad healthcare market. For example, Pure Cardiology is a user-centric membership plan for all cardiology-related services, the Surgery Center of Oklahoma offers a menu-style list of surgeries they perform with up-front pricing, and the East River Medical Imaging Center in New York is an independently owned and operated center for scans including MRI, CT, PET, and X-rays.

In primary care, there has been a recent increase in the number of direct care clinics around the US, or doctors that are offering traditional primary care through an independent clinic without ties to large hospital networks and insurance companies. To support these practices, companies like Hint Health (onboarding and billing platform), Spruce (patient communication), and Elation (clinical electronic health records) are building the infrastructure to power a direct-care driven healthcare system. Technology enables new user experiences for quick onboarding and communication, but other offline examples of lowering friction are walk-in urgent care centers, such as CityMD and GoHealth. These clinics provide immediate medical care for a range of acute conditions that provide underserved patients with faster response times. Walk-in urgent care clinics started popping up in the 70s but gained more traction in the 90s and now service over 160 million visits annually.

These examples suggest that healthcare is in some respects being unbundled.

Why does this matter? One lever to getting momentum in an unbundled world is that a lower price point - a common denominator across many of these companies - compared to traditional bundled options, will allow more people to access healthcare. Besides just lowering costs, unbundling can open access in other ways, including minimizing the enormous friction that exists in legacy healthcare systems. If healthcare services continue the trend of “unbundling” in some of these ways, cost-efficiency will become much more of a priority where the fastest, cheapest, and most reliable services should win. More fundamentally, unbundling could create the opportunity for each of these distinct services to become independent nodes in a larger network in a bottoms up manner (which would make room for business model innovation - the subject of my next blog post).

For now, these observations seem to indicate some unbundling of the existing large, monolithic systems in healthcare towards a more open, local, independent and transparent model, with control residing with individual users. And ultimately, this could change the way healthcare is delivered to consumers.

USV’s Back to School Reading List

It's been a busy summer of reading for us here at USV. Since we often receive requests about what books people are reading, we polled the USV team internally to find out what picks and recommendations they each wanted to share with all of you. 

As summer is quickly coming to an end, we hope you can consider this your "Back to School" reading list from the team at USV. (If you have other book suggestions or recommendations, please leave them in the comments!)

Our Top Fiction Picks:

The Little House by Virginia Lee Burton
Recommended by: Jed Schmidt
It's a popular children's book from the 1940s about a house in the countryside that slowly gets engulfed by the growing city. It's pretty poignant, especially considering its intended audience.”

Lexicon by Max Barry
Recommended by: Dani Grant
The book is a thriller. It's summer. Enjoy.”

Sweetbitter by Stephanie Danler
Recommended by: Lauren Maz
“If you've ever worked in any restaurant, anywhere you might find this engaging.”

The Man in the Basement by Walker Mosley
Recommended by: Nick Grossman
“So good”

Less by Andrew Sean Greer
Recommended by: Nick Grossman
“Also Fantastic”

Another Brooklyn by Jacqueline Woodson
Recommended by: Lauren Maz
“It’s a lovely quick read”

A Gentleman in Moscow by Amor Towles
Recommended by: Lauren Maz 

You Should Come With Me Now by M John Harrison
Recommended by: Andy Weissman
“Collection of genre subversive short stories and flash fiction.”

Lincoln in the Bardo by George Saunders
Recommended by: Andy Weissman
His tweet speaks for itself
Autonomous by Annalee Newitz
Recommended by: Albert Wenger

Non-Fiction Recommendations:

When Breath Becomes Air by Paul Kalinithi}
Recommended by: Naomi Shah
“One of my favorite books. It illustrates an interesting dynamic when the doctor becomes the patient. Having had a family member with cancer, this book hit home for me.”

Bad Blood  by John Carreyrou
Recommended by: Zach Goldstein
“In a world where technology companies can be glorified, this is the story -- from the perspective of the WSJ reporter who broke the news -- of a generation changing business called Theranos gone terribly wrong, with lessons about business, life, and human nature intertwined.”

Fantasyland by Kurt Andersen
Recommended by: Albert Wenger

Savage Inequalities by Jonathan Kozol
Recommended by: Bethany Crystal
“This book takes a deep dive look at the education system in the U.S. and the vast disparity of funding gaps and opportunities seen for children in urban areas in the 1990s. It's an examination of how the design of this system continues to compound the impact of segregation of predominantly White communities from Black, Latinx, and Asian communities.”

Cable Cowboy by Mark Robichaux
Recommended by: Andy Weissman
“A well written, fast paced, history of the entrepreneurs who invented the cable tv business.”

The Founder's Dilemma by Noam Wasserman
Recommended by: Jennifer Greenberg
“This book walks you through all of the decisions that founders have to make by showing you actual startups and their journeys”

Ghost In The Wires by Kevin Mitnick
Recommended by: Dani Grant
“This is the book I read that first interested me in computer security. If I hadn’t read this book, I wouldn’t have applied to join Cloudflare and then wouldn’t have met the USV network. Oh and one tip: it’s great as an audiobook too.”

Slugfest by Reed Tucker
Recommended by: Fred Wilson
"Still reading it, but enjoying it so far"

Analyzing Tools Used by Our Network

This summer at Union Square Ventures, Jennifer Greenberg joined us as our summer intern. During her time at USV, she worked on many projects, and one of them was analyzing tools from across our portfolio companies. As part of this process involved grouping and sorting tools, she came away with a few observations from her project. Read her post below.

Hi, I’m Jennifer Greenberg, a current computer science student who had the pleasure of interning at USV this summer to support the Network Team.

Through my community management work in Slack as well as in attending USV network events, I noticed that one of the most common questions that comes up in the network is what tools are used by other companies.

To embark on this research project, we gathered over 450 software tools used across 64 companies, representing approximately 86% of our active portfolio. We began by tagging each tool based on department, the type of tool, and then explored how company size impacted usage.

Below are some observations and trends that we discovered.

Types of Tools Reported

As you can see below, developer tools ranked #1 among the categories of reported tools for this project, with 164 total tools reported. We also saw large numbers of tools used for communication, business intelligence, and design   

Most Popular Tools

The top reported tools were G Suite, Slack, GitHub, Jira, Sketch, Salesforce, DropBox, AWS, Excel, and InVision. These top tools (in particular G Suite, Slack, and GitHub) were used across multiple departments or tended to skew toward developer tools.

Developer Tools in the Network

The chart below show the most popular developer tools overall in the network. After Github and Jira, there was a drop off which may be because there are many free/open source options, causing individuals and companies to explore and favor different ones.

Size Breakouts

Company size creates a stratification among developer, hiring, and communication tools. As an example of how this plays out in engineering, you can see that all of the size brackets use either Github or Jira, but after that, the differentiation develops. For example, smaller companies are less likely to use security software. On the communication side, smaller companies are more likely to use internal tools like Calendly, GoToMeeting, and Rocketchat, whereas external tools including Docusign, Facebook, Medium, Sendgrid, and Twitter persisted across all size brackets.


As for hiring tools, LinkedIn, Lever, and Greenhouse are the most popular in the network, but these are primarily only used by companies with more than 30 employees. Tools like Guru, Textio, DiscoverOrg, and Checkr were used with companies larger than 50 employees, while Jobbatical, Entelo, Jazz, and AngelList were used by our smaller companies.


Market Dominant Tools by Category

While AWS still appears to be the favored cloud platform used among our portfolio, Google Cloud in particular appears to be gaining more traction. As you can see below, ¼ of companies are using either Google Cloud or Microsoft Azure.

Other tools with dominant market share include Slack and Salesforce. While 92% of our companies report using Slack, some companies opted to use Hipchat (recently acquired by Slack), Beekeeper, and Rocket Chat.

For sales management CRMs, 42% of our companies use Salesforce, however, some used alternatives like Google Streak, Insightly, or It does not seem as though size played a role in a company's choice in any of these circumstances.

Closing Thoughts

One of the most fun parts about working on this project was learning about new tools used in the network. In fact, 52% of the tools reported were used exclusively by one company in our portfolio, which exposed a few new tools like Jell and Perdoo (both workflow management tools) as well as Sapling (an HR platform built for G Suite). It was also great to see how 78% of our portfolio companies use tools built by our portfolio companies (including Code Climate and Cloudflare, among others.)

While the average company in our portfolio is using 7 tools, one of our larger companies reported using 77 different tools, which made me realize that transitioning between tools may be a source of growing pains for startups. As our portfolio continues to expand, we hope that by aggregating and sharing this list internally (we’re even building a Slack Bot for our network to search these tools), we can make it easier to choose the best tool for each use case.

The Distributed Computing Update

One of the applications of cryptocurrency we continue to be excited about is distributed computing.

Before crypto, my laptop couldn’t pay a stranger’s idle server as a thank you for running a machine learning program. Cryptocurrencies finally give us the ability to make machine-to-machine payments to compensate participating nodes for running tasks.

In June I wrote an overview of the types of compute projects we were seeing. It’s been two months, but this is a space that moves fast and I wanted to keep sharing what we’ve been learning. Here goes.

Siloed networks vs an open protocol

There are two ways distributed compute could play out. In one model of the world, there is a dominant distributed compute protocol that creates a shared network of machines that anyone can build interfaces and clients on top of. Think of this like Heroku and EC2: both of them run on AWS servers, but they offer interfaces with drastically different experiences that cater to different audiences.

In the other model of the world, there are a few dominant compute projects that each have their own network of machines.

Both worlds allow for there to be coexisting projects that serve different audiences, but in one version of the world, the projects are clients on top of the same shared resource pool, and in the other, they all run their own independent networks. It is possible that these two models co-exist, but I think that is unlikely because of network effects. If given the opportunity, projects may opt to plug into an existing network of machines rather than build their own because having access to more CPU gives them better quality of service for their customers on day one than if they had to start from scratch.

We are seeing attempts at both. SONM is one project trying to build the shared resource layer. Another is Distributed Compute Protocol (DCP), built by Distributed Compute Labs. Most other projects are currently building out their own networks, though with open protocols there is really nothing stopping anyone from building alternative interfaces to any of these projects. We may see projects start as their own system and organically grow to be just one of the clients on top of their now shared resource layer. I am pretty excited about the possibility of a shared compute layer and about the teams and projects that are trying to build it. 

Typically the opinion of clients built on top of an open protocol is that they are brutally competitive because it is easy for a user to move from one client to another. Think of this like early Twitter where lots of people were building Twitter clients. It was easy for a user to move from one Twitter client to another so being in the Twitter client business was hard and competitive. This may be different with computing where the interface that the client exposes is the product. If it turns out to be the case that the client APIs are different from one another, because developers integrate them into their source code and CI/CD workflows, the clients could be incredibly sticky even if they all effectively expose the same backend. I think that is an important feature of compute that will even further incentivize projects to contribute to and build on a shared resource pool.

Token Questions

One question we have been thinking about is which tokens will be used by developers versus which tokens will be used by end users. That is: if a user interacts with a dapp that runs code on a distributed compute network, does the user pay the dapp in the same token that the dapp pays the compute service?

Right now the trending answer in compute services is no. Akash, Render, Perlin, Enigma and SONM are some of the compute projects that have their own transactional token. This follows the same model as IPFS/Filecoin where users will presumably pay dapps in whatever the major consumer-facing currency is (right now it is seemingly ETH or BTC) and dapps will behind the scenes exchange that token for the tokens they need to provide the service.

Hypernet and Truebit, on the other hand, are two compute projects with two-token models. In Truebit, for example, buyers can pay for the service in ETH, and the Truebit TRU token is just used for the protocol-specific functions of staking and dispute resolution. This matches a pattern we are seeing this year with infrastructure projects like The Graph and Augur that use the main consumer currency for transactions, and their own token only for governance, staking and dispute resolution. I predict we will see more projects change to the two-token model because it allows the price of governance to increase as the network grows, but doesn’t increase the price of the service with it.

The EC2 Model vs. The Lambda Model

In the existing web2 world, there are two main types of compute services: in the EC2 model, developers are provided an environment to run and host services, and in the Lambda model, developers write functions that can be invoked on demand.

The distributed computing projects break out into these two categories as well: one is like Lambda (or like Cloudflare Workers 😉) , the user writes a script, and the project runs it on participating machines. The other approach is the EC2 approach or the “someone else’s computer” approach: the user gets matched with someone on the network and can run a container on that someone's machine.

Note that the Lambda approach isn’t quite Lambda yet - machines in Lambda-like distributed networks don’t store all of the functions ever pushed to them and invoke them on demand. Instead, these networks are for running offline and async scripts for use cases such as scientific computation or rendering graphics. As latency improves, we can see these becoming more like serverless compute over time.

The ecosystem needs both models: hosting a dapp front end requires a persistent host, and running one-off computations is better on a serverless-like platform.

Two projects working on hosting platforms are Akash and DADI. Akash actually looks very much like traditional compute services from the end user’s point of view - developers manage containers on Akash-deployed machines in a Kubernetes cluster that can be federated across machines on the Akash network. (Not coincidentally, Akash is founded by Greg Osuri who is also a contributor to Federated Kubernetes). If you’re curious to try Akash, they recently launched a testnet.

Two projects working on the serverless platforms are Ankr and DCP.

Oh, the devices you’ll go!

The thing that distributed serverless compute projects can do that feels unique to cryptocurrency-based distributed computing networks is that they can run code on strangers’ phones and laptops because they don’t need to persist the compute environment beyond running one small script at a time.

The idea here is that these projects can pool together all of the unused end user CPU to form a giant super computer that is cheaper than what is available on the cloud compute market today.

[Side tangent on pricing: The main argument here is that distributed networks will be cheaper because they do not have to pay for physical space and the hardware capex cost has already been committed. However, as Mario @ Placeholder pointed out to me, cloud compute pricing is already racing to the bottom and if distributed services come along and undercut the main players, cloud providers can presumably come all the way down to just above maintenance cost and stay competitive.]

I am very excited about projects that can provide access to high power compute environments by pooling together available CPU on end-user devices.

There are three big challenges with running code on end-user devices. The first is convincing enough individuals to participate, which we covered in our previous post.

The second is that end-user devices are relatively low power. To counter this, we are seeing projects building in parallelization to run code simultaneously across multiple machines at once. Ankr leaves it up to the user to package their code into chunks and submit them separately to the network, where they will be assigned by a job scheduler to different machines. DCP auto-magically distributes an application’s subtasks across machines in the form of JavaScript objects that execute in web workers (DCP also cleverly uses WebGL to tap into the GPUs on end-user devices for an additional boost).

The third challenge is that end-user devices are untrusted. There has been a lot of recent momentum in utilizing SGX, a trusted hardware environment built into Intel chips, even since our last post in June. Since then, Enigma released a testnet utilizing SGX for compute, Golem released Graphene-ng to help developers write SGX-enabled code, and Oasis Labs raised $45M led by a16z’s new crypto fund to build SGX-enabled distributed compute. The top 3 laptop makers: HP, Lenovo and Dell support SGX. MacBooks have SGX-enabled chips, but the BIOS hasn’t been configured to expose that functionality up to the operating system. When Apple adds SGX, the top 4 global laptop brands will all have built in support for SGX-enabled computing. I am a big supporter of the SGX approach because it is fairly secure and accessible on consumer laptops.

Besides SGX, another way distributed compute protocols can verify computations is with dispute resolution. Truebit is one of the compute projects with a dispute resolution protocol, which they call a “verification game”. In it, a verifier stakes TRU tokens to challenge the result of a computation. In Truebit’s dispute resolution game, the solver’s state is hashed at each time step of running the program (actually, any given instruction might not be executable within Ethereums's gas limit, so TrueBit breaks down each instruction into 12 substeps). Then the verifier queries those hashed states using binary search to find the exact instruction at which things went awry. The disputed step or substep is then run on Ethereum for the final outcome. Whichever side is wrong loses their staked tokens, which are paid out to the winning side.

Where on the stack does compute fit?

One open question is whether compute services will end up being a layer 1 or layer 2 solution. That is: will the next major blockchain include compute as a built-in service, or will compute always be run off-chain.

The reason why compute is done off-chain now is because the predominant blockchains available for use are either Bitcoin: limited scripting language or Ethereum: compute is expensive and slow. There could very well be a future in which a layer 1 blockchain is able to bake in compute in a way that doesn’t require every node in the network to run the same computations, which would make it cheaper and faster. Perlin is one project attempting to build this, though even in Perlin, compute services are implemented as a side chain of the main Perlin base chain.

Most projects are either building side chains to existing blockchains, or completely off-chain networks that are independent from existing base chains. Render is one example of the first approach - Render is implemented as an Ethereum smart contract that interfaces with the Render network. Akash is an example of the latter: it is a separate network entirely.

I tend to like light, horizontal protocols that can be layered on one another rather than a super-protocol blockchain that can do everything. That is how the internet works now - small protocols that layer on top of one another (SMTP > STARTTLS > TCP > IP). What it allows for is reusable modules (both QUIC and DNS can use UDP without there needing to be changes to UDP to support that) and the ability to easily swap out and upgrade layers (HTTP can be swapped with SPDY or upgraded from HTTP 1.1 to HTTP 2.0 without making changes to the layers below it).

Geographical Market

The last thing to say here is that one potentially very smart thing we are seeing is projects focusing their approach on one geographical market. DCP, for example, is starting by providing compute to Canadian universities and labs (though through the process they have picked up a lot of interest from outside of Canada as well). Ankr, for example, is putting extra effort on reaching the Chinese computing market where demand for compute is skyrocketing (Aliyun’s revenue grew 104% year over year) and AWS doesn’t have too much of a stronghold (though Aliyun does). We think these targeted approaches could play out well.


It is still early days and there are a lot of unknowns, but we are optimistic about what could be. If you are building interesting projects in this space, we’d love to hear from you. Reach out: I’m [email protected]

Jed @ USV

Hey all, I've recently joined USV as the firm's very first Developer in Residence. Over the next year I'll be porting USV's thinking around networks into code, and writing the APIs, apps, and bots that will help folks at our 70+ portfolio companies communicate and collaborate with each other.

Having an in-house developer is still a bit rare in venture capital, but it feels like a natural fit for USV. Building on its early success investing in businesses that leverage network effects, USV brought this approach to its own network in 2010, launching an effort to facilitate interaction among network members through events and introductions. And now my job is to grow this platform online, to help reach and connect a greater share of folks in our increasingly diverse portfolio.

So far, my work has been roughly split between the back- and front- ends. On the backend, I'm writing code (mostly JavaScript running on AWS Lambda) to aggregate and consolidate several years of data from the various services we use into a unified view. This will help USV's network team figure out how to allocate our time and attention to better serve our portfolio companies. On the frontend, I'm building tools to connect fellow portfolio members where they hang out, which for us these days means Slack. We've built a Slack bot that helps members discover each other through similar interests and tools, answering questions like "Who is working in sales at USV companies with 50-100 people?" and "What design tools are popular among other developers in the network?"

Helping like-minded folks discover and learn from each other was one of my favorite things in building BrooklynJS, my favorite community here in the city. To be able to do the same for USV is a great gig, and really, second only to my other gig, moonlighting as a bass for Brooklyn's favorite-slash-only barbershop quartet.

If you’re building a similar platform and/or have ideas about what kinds of interaction mechanics work well for community building, please drop me a line!

An Overview of Blockchain-Based Universal Basic Income Projects

Universal Basic Income (UBI) is the idea that citizens receive a regular, unconditional stipend that helps them cover their cost of living. Previous UBI experiments have shown to reduce hospitalization, crime and poverty rates. Richard Nixon, Thomas Paine, Martin Luther King Jr. and Milton Friedman were all vocal proponents of UBI.

UBI has traditionally been imagined as a government subsidy that would put money back into the economy by giving it directly to people (as opposed to quantitative easing where the Federal Reserve puts money back into the economy through banks).

The development of cryptocurrency, however, now gives us a way to implement UBI in a global, trustless and democratic way without the need for a government to implement it.

Recently there has been an emergence of a handful of blockchain-based UBI projects. They are all very early. Most of them do not yet have a public product, but a few do, if you’re curious to try some out, a few you can check out are Mannabase:


and Solidar (implemented as a chatbot on FB messenger).

We are intrigued by this possibility and are wondering about some key issues, such as the complexities around issuing new currencies and preventing fraudulent accounts.

Where does the money come from?

When blockchain projects implement UBI, where does the initial money come from?

The majority of the UBI blockchain projects issue their own currency in the form of tokens. That is, instead of recirculating existing money in the economy, they generate new value by minting a new currency. The challenge is that while the idea behind UBI is to provide real income that can be used for paying for things like rent, tuition and groceries, newly invented currencies are initially worthless until someone accepts them. It is up to each UBI project to make their currency worth something.

Projects do this by building an economy around the currency where people can exchange and use their tokens to buy goods and services. Nick calls this building a ‘Minimum Viable Economy’.

Building a Minimum Viable Economy: Vendors & Merchants

The idea behind a Minimum Viable Economy is to build enough of an ecosystem around a token so that its holders can use it to buy goods and services or exchange it to other currencies.

For this to happen, the project needs to incentivize merchants and vendors to accept the token as a form of payment.

SwiftDemand is probably the UBI project with the most developed marketplace so far. Their hope is to seed the marketplace with vendors that are participants in their UBI community. Anyone in the community can submit something to sell:

And then anyone in the community can buy those things using the Swift token:

projectUBU (beta) is building tools for vendors to be able to easily add support for their UBU token. Enumivo (pre-beta) is building their own blockchain (a fork of EOS) with the goal of developers building dapps that accept their token, $UBI.

It is easier to convince vendors to accept a token if there are a lot of people that hold the token. A good analogy for this is the credit card: even though vendors dislike credit cards because they are expensive and require extra in-store hardware, they are incentivized to accept them because so many people have them.

To seed this network effect, many UBI projects have referral programs to reward people who bring in new users. projectUBU, for example, rewards 1,000 UBUs to the referrer and 500 UBUs to the referee per referral.

Some projects, instead of doing a one-time bonus, continue to award the referral bonus as long as the referred person stays in the network. Frink (beta), for example, plans to indefinitely payout an additional 10% to referrers, and Mannabase plans to payout an additional 100% to referrers for one year. The idea is to incentivize people to refer “high quality users” that will stay in the network for a long time. An interesting question is whether a high referral bonus will increase the incentive and potential for referral fraud.

These referral programs are often set to expire when the network grows to a desired size. Solidar’s program, for example, is scheduled to reduce the bonus by 50% when the network reaches 15,000 users and then again every time the network size doubles.

Building a Minimum Viable Economy: Monetary Policy

Projects also need to incentivize people to spend their tokens. UBI projects can build in monetary policy that makes it more attractive for token holders to spend the tokens than to hold them.

There are two ways to do this: demurrage (some amount of held currency automatically dissipates) and by growing the money supply (so that each held token is now worth less). Both accomplish the same goal of incentivizing token holders to spend their currency, otherwise their held currency will lose some of its value.

projectUBU is one of the projects utilizing demurrage: 1% of all UBU wallet balances dissipate every year. Circles is one of the projects planning to mint more currency: they plan to grow their money supply at a 5% annual rate. The most dramatic of these programs is Solidar, which has an annual 20% demurrage rate.

Another way projects incentivize people to spend their tokens is by capping the amount of tokens any account can hold at one time. In order to receive more tokens, participants need to withdraw or spend the tokens they’ve already received. SwiftDemand, for example, only allows accounts to hold 7 days of unclaimed income at a time.

Building a Minimum Viable Economy: Liquidity

Another way to create value in tokens is to provide liquidity - aka the ability for a token holder to exchange the token for another currency, usually fiat, like USD. 

For there to be liquidity, there needs to be someone who wants to buy tokens from those that hold it. 

One project called Big Foundation (beta) is seeding liquidity by paying people a bonus for buying the token.

Greshm (pre-beta) holds a reserve of USD and issues currency called XGD backed by that USD reserve. (Note that they are built on their own system and not on blockchain). That provides initial participants and vendors with a source of liquidity - they can cash out and receive an equal amount of USD for their XGD. Greshm plans to maintain a 1:1 peg to the USD at first, and then increase the ratio of XGD to USD over time. This will allow them to put new money into circulation. (This model exists in the traditional US economy where federal banks can create new money by lending out money they don’t have in reserve up to a certain lended_money:money_in_reserve ratio.)  

Another interesting approach here is Democracy Earth’s distribution program. Because their currency has immediate utility as a vote, there are more likely to be buyers of it. Democracy Earth (beta) is a governance platform, and buying currency can mean buying power. The caveat is that organizations built on Democracy Earth can set their constitutional smart contracts to limit only one vote per person per issue, which inhibits the ability for participants to effectively buy votes.

Identity Verification & Anti-Fraud

Before a UBI project can hand out tokens, they first need to verify that each participant is a real person, and that each person is limited to a single account. This prevents cheating via ‘Sybil attack’ where a user creates multiple identities that all trust and validate each other in a closed system. If every user could create multiple accounts to increase the amount of income they received, it would dissolve the public trust in the value of the currency, and depreciate its worth. It would also undermine the spirit of the project where in everyone gets the same amount.

There are two main ways that UBI projects are solving this: voting and social trust.

The first way is allowing members of the community to vote to verify a new participant. On Democracy Earth, for example, new participants have to go through a validation process with other previously validated community members in order to be able to join the network. (They actually plan to have every participant repeat this process periodically in order to prevent abandoned accounts).

The second way is by relying on trust relationships from the real world. Circles (pre-beta) does this in an interesting way: On Circles, each new participant is issued UBI payouts in their own personal currency. That currency is not worth anything because no one agrees to exchange it yet. To make their account balance worth something, Circles participants need to trust each others currencies by being willing to exchange them. From the Circles documentation: "The value of a specific personal currency is a measure of how many other accounts trust it. This means that users who are new to the system and don’t have many trusted relationships have a less valuable currency than someone who is well-established in the network. It also means that the currency of new users gets more valuable over time as they create more trust relationships."

Enumivo plans to do a combination of the social graph and voting solutions. People who want to join Enumivo will have to find someone already in the community to sponsor them. To sponsor someone, a community member stakes 200 tokens (10 weeks worth) and then other community members have 30 days to vote on them.

There are also standalone identity projects like uPort and Civic that future UBI projects could potentially leverage. Generally we are very interested in learning more about self-sovereign identity projects that could enable decentralized programs like UBI.

Are These Projects Sustainable?

There are two ways most UBI projects fund their development: by holding a percentage of their tokens (most UBI projects do this), and by collecting transaction fees (some UBI projects do this).

What I like about these revenue sources is that they align the core team’s interest with their users’ interests. The better the core team grows the network and token economy, the more their tokens are worth, and the more transactions there will be to collect fees on.

Wrapping Up

One of the applications of blockchain that we are very excited about is UBI, and we hope to keep learning about how different projects are implementing it. If you’re working on something in this space, we’d love to hear from you. Reach out, I’m [email protected]

USV Intern Day

I know my favorite day of the year should be our USV CEO Summit. But it’s not. It’s actually USV Intern Day

On this day, which we now host once a year in both NYC and SF, we bring together interns from dozens of portfolio companies and invite them to learn about all of the companies in our portfolio. For me and Lauren, this day feels a little bit like summer camp -- we sport swag from our portfolio companies and shuttle interns around the city all day.

Like all of our best programming at USV, this idea didn’t originate at USV, but it came from our network. In 2016, the talent and people team at Meetup sent this email, asking:

“Are there any USV events geared towards interns? Has USV thought about doing some type of event/crawl for interns to visit different offices in the portfolio? We thought it'd be great to give our interns an opportunity to interact with interns at other portfolio companies.”

We jumped on it immediately and rallied participation from more than a dozen NYC companies. For participating interns, we organized a multi-stop event with two tracks (technical and business) featuring short talks, office tours, and job advice from the incredible leaders in the network. Some companies handed out free swag to all attendees. Meetup concluded the day with a pizza party on their incredible roof deck.

But I don’t like Intern Day because of the pizza and the swag. I like it because of the deep collaboration and sense of community that it represents among our portfolio.

This year, we had about 75 interns between our San Francisco Bay Area companies and our NYC companies. Product and operations leaders from Skillshare and Code Climate each took an hour to introduce these students to the complexities of their business models and careers in tech. Shippo in San Francisco facilitated an entire panel discussion with robust Q&A. Clarifai taught a group of 40 students how they think about AI and machine learning. Matt Blumberg, CEO of Return Path, spent the last hour of his busy workday sharing candid advice he wished someone had taught him before entering the workforce.

When kicking off the day, we shared two stories with the intern class this year. The first was the story of how Andrew Sutherland, at age 15, started a flashcard app for himself while studying for a high school French test that eventually became the company and brand we all know today at Quizlet. The second is how, in their first first job after school, Meetup CEO Scott Heiferman met his now-co-founder Brendan McGovern when they sat next to each other on the very first orientation day at Sony.

One of the things we are so privileged to see at USV are the origin stories of entrepreneurs all over the world. We hope that, even by bringing together interns for one day out of the year during their summer, we’ll spark some inspiration or ideas that will come back our way several years down the road.

Investing In Token Focused Funds

At USV, we have been active in the blockchain/crypto sector since 2011. At this time, we have direct investments in nine companies/projects in the crypto space and one exited investment. Companies in the blockchain/crypto sector make up about 15% of our active portfolio and closer to 25% of our recent investments.

But the venture capital fund model is not optimized for investing in the blockchain/crypto sector. Blockchain/crypto companies/projects often finance and monetize via tokens which can become liquid quickly and thus we can end up holding highly liquid and volatile positions which is not something we have traditionally done. And because USV operates under the venture capital exemption to Dodd-Frank, we are limited to 20% of our holdings at cost in “non-qualifying” investments, which include tokens.

So we have sought out other relationships in the sector that can allow us to get broad exposure to the most interesting companies and projects. Our most recent investment, which closed last week, is just that. USV became investors in Multicoin Capital, a fund that is focused exclusively on the blockchain/crypto/token sector. There are a lot of smart people thinking about and analyzing this emerging sector, but there aren’t many who are doing it so publicly and conversationally as the Multicoin team. Their blog posts are here and their tweets are here: (Kyle, Tushar) . Their opinions are often controversial and contrarian. You can make a lot of money by being right about something most people think is incorrect. So at USV, we appreciate and value original thinking.

Multicoin often talks about “venture capital economics with public market liquidity.” The token sector offers both, and that is partially what has caused USV some challenges making this sector work in our current fund structure. Multicoin is structured in a way that gives them and their investors the benefit of both. We are excited to be investors in Multicoin.

Multicoin joins a roster of other blockchain/crypto/token funds that USV has invested in over the last 18 months. That list includes Polychain Capital, Metastable Capital, Blocktower Capital, Arianna Simpson’s new fund Autonomous Partners, and Placeholder. Placeholder, like USV, uses a venture capital fund model for its investing. All of the other funds use more of a hedge fund model.

This portfolio of token funds gives USV a much broader reach across the blockchain/crypto/token sector than we would be able to get directly on our own and we also benefit enormously from the dialog and information sharing that exists across this network of investors.

USV has not become, and has no plans to become, a fund of funds. But the blockchain/crypto/token sector has some unique challenges for us and others in the venture capital business and we have taken a network approach to solving them, at least for now. 

We have and will continue to invest directly in both companies and projects and both equity and tokens in this space, often in a syndicate that includes one or more of these token funds. It is an exciting time to be investing in blockchain/crypto/tokens and we are fortunate to be able to do it along with some of the best investors in the sector.

An Overview Of The Distributed Computing Landscape

People have been trying to build distributed compute networks since the 1990’s; In 1996, GIMPS used distributed compute to search for prime numbers and in 1999, [email protected] used volunteers’ compute power to search for extraterrestrial life.

Now 25 years later, the final pieces seem to be in place. Cryptocurrency makes possible machine-to-machine payments, which allows participants to get compensated for contributing CPU. Fields such as machine learning, 3D simulation and biological computation are driving up demand for compute resources.

We’ve been looking at distributed computing projects and wanted to share how different projects are tackling growing the number of machines connected to the network and isolating tasks from the compute nodes they run on.

Below are our early findings. We hope they are useful, and let us know if you have any feedback.

Approaches To Growing The Network

Metcalfe’s law applies to compute networks: the more machines there are on the network, the more likely a machine will be available to accept a new task when needed.

Growing a compute network is difficult to do, especially as the space is increasingly crowded. To clarify - the issue isn’t that people already have installed something and don’t want to install something else, but rather that there is a lot of noise for a project to break through.

Here are four interesting approaches we are seeing:

Approach #1: Make it easy for anyone to participate in the network. One example of this is KingsDS (pre-beta). To join, all you need to do is visit a URL in the browser and let the tab run in the background.

Approach #2: Help other applications get compensated for pooling their own users’ resources. An example of this is  FREEDcoin (pre-beta). They offer an SDK for game developers. When players launch games running the FREEDcoin SDK, they are given the opportunity to contribute their CPU in return for in-game prizes. It’s a win-win-win: FREEDcoin gets to add high-power gaming PC’s to their network, game developers can monetize their games without showing ads, and players have the opportunity to earn virtual prizes.

Approach #3: Build the client so that each node can both submit and complete tasks. Golem’s (beta) client can be used to submit tasks and to compute them. That means each one of their end users can also easily become a compute node. This helps them grow both sides of their network evenly.

Approach #4: The last approach is to be the supplier of compute resources for other computing projects. One example is SONM (beta), a project trying to help other compute networks scale up quickly. With SONM’s open marketplace, machines can advertise how much RAM, CPU and GPU they have available in a standardized format. Any project using SONM can then search the entire SONM network for a machine with available resources.

Approaches To Isolating Tasks From Host Machines

One challenge is ensuring that tasks cannot read or modify memory of their host machine and vice versa. If multiple tasks are running simultaneously on a machine, it’s important that they are isolated from each other as well.

It’s a tough challenge to keep data private; even though SONM runs all tasks in Docker containers, they also have partners that run nodes sign NDA’s.  Most projects are relying on existing container runtimes like Docker to satisfy this requirement. Makes total sense - who wants to reinvent wheels. However, there are two projects in this space that are doing something unique and are worth calling out.

Enigma (pre-beta) is designing what they call “secret contracts” - these are compute nodes much like smart contracts but because every piece of data is split across multiple nodes working on the same compute task, no single node can read any data. They do this using a cryptographic method developed in the 1980’s called multi-party computation. Enigma is building out their own chain that will be able to do the storage and compute.

Keep (pre-beta) is another project taking a similar approach. They are also using multi-party computation to shard encrypted data to perform computation without the compute nodes being able to read any inputted data. With Keep, the storage and compute of private data happens inside clusters and the output gets published on the blockchain.

One Last Thought: Narrow Vs. Broad Use Cases

There are two approaches one could take for a distributed computing project: build a general compute tool that could accept any workload or accept only a narrow range of tasks.

Most of USV’s portfolio companies started by doing one thing, and doing that one thing allowed them to grow and build a network and a platform around that one thing. (e.g. Cloudflare, Stash, Carta, etc.)

I tend to think that the same pattern will work well for compute networks: starting with one narrow use case (such as training machine learning models, rendering 3D shapes, and folding proteins) will help a project move quickly and over time grow into other compute areas.

Albert likens this to WeChat’s growth: WeChat started with chat and the success of chat allowed them to grow their network so that they could build other applications like payments, ecommerce and gaming, and now WeChat is a general use tool.

There’s a question around what is the right use case to start with. There seem to be two paths: one is starting with training machine learning tasks (machine learning is one of the drivers for increased demand for computing resources). The other is starting with a use case like 3D rendering or academic/scientific computation where there is no overhead of private data to protect.

Wrapping Up

This space is early, but an exciting prospect. Not only will greater competition in compute providers drive down prices and fuel innovation, but there may be a new class of applications (such as VR and autonomous vehicles) that may only be made possible when distributed computing will be hundreds of milliseconds closer to end devices than us-west-2.

If you have ideas or projects you’re working on, we want to hear from you. Reach out, I’m [email protected].

Dani @ USV

Hi everyone. I just joined USV as an analyst and am very grateful to the USV team for the opportunity to learn from them. I’m coming from Cloudflare, a USV portfolio company, where I worked for Dane on the Product Strategy team. I was there for 3.5 years and worked with the team there to launch a lot of fun projects like, and Cloudflare's partnership with F-Root.

I’ve been at USV for 2 weeks. Here are some things I’ve learned so far:

1. Investors at USV are in a day-long meeting every Monday so the worst time to cold email a USV partner is on a Monday.

2. One winning presentation format is to start with growth numbers. A lot of company presentations start with describing the product first, but nothing grabs investors’ attention like proof in data.

3. This one I learned from Assif: Thinking of tokens as an asset class is missing the picture. If tokens are successful, everything will be tokenized (sports teams, companies, real estate, art, etc), and more people than ever will own pieces of tokenized assets.

4. Some interesting blockchain projects are starting with Android-first because it’s easier to grow a network in places where the few tokens you get in return for participating in a network are more meaningful.

I’ll be continuing to share what I’m learning on this blog - stay tuned for an overview of distributed compute projects later this week. If you’re working on anything interesting related to our new thesis, reach out. I’m [email protected]  Excited to start the Union Square adVentures.