Packy McCormick, the founder of Not Boring, a newsletter turned VC fund, recently announced his third venture fund, Not Boring III. It’s a $30 million fund that will, as he describes, “invest in hard startups”. Packy defines hard startups as a point in the market — an inflection point of sorts — where technology goes from the investigative discovery phase into the deployment phase. His argument is that for VC to continue to generate historic returns, they have to move earlier in the innovation cycle — into the hard territory. To steal the visual from his post:
Since I’m sure Packy would agree all startups are hard, let’s categorize startups as either hard or less-hard — nothing is easy. However, this dichotomy of hard versus less-hard startups manifests in more than just where on the innovation curve you build your company. Even at a single point on the curve, there are hard and less-hard companies to build — or better yet, hard and less-hard problems to solve.
The Two Problem Spaces
Over the last 3 years, we’ve seen a few technology hype cycles come and go. But no transition between cycles has been starker than the crypto to Chat GPT excitement shift we’ve seen over the last few months. This tweet sums it up nicely:
Looking at the rapid shift between these hype-cycles, we can see a common mistake founders make — following the temptation of a trendy space. But the trendiness of a space can actually be a disadvantage.
From early in the pandemic through 2022, crypto, or web 3 as it became optimistically named, exploded. NFTs, tokens, new exchanges — the crypto world was exploding with new companies created daily. Superbowl ads starring Matt Damon should have hinted to us we were near the peak, but before long the new asset class came crashing back down to earth.
Crypto’s hype cycle was quickly replaced by Open AI’s release of Chat GPT 3.5 into public beta. The world went crazy with the new open API and some reports claim that 50% of the startups in the current Y Combinator batch are using Chat GPT. The flock of builders into this new space is nothing to be scoffed at, many large companies will be built using Chat GPT. But as we’ll discuss, when the masses gravitate towards a new space, they tend to address the less-hard problems, not the hard ones.
Apps vs Infra
Most software spaces can be divided in two: the infrastructure and application layers. The infrastructure is the piping — the architecture that enables us to build the application layer. Deep technical innovation tends to happen on the infrastructure level, while experiential enhancements come on the application side. Dani Grant and Nick Grossman from Union Square Ventures wrote a great post on this back in 2018, about how hype cycles move between application and infrastructure investments in order to progress. They explain that it’s not sequentially infrastructure first then applications, but a constant shift between the two. To steal a visual from their post:
At any point in time, either apps or infrastructure can be built. But, depending on the point, one is harder and one is less-hard. We’re seeing that play out today in crypto and AI.
The greatest hurdle in generative AI is an infrastructure challenge. To build an engine that, when prompted, is capable of generating human-like content; something indistinguishable from what a human would have created, only done by algorithm. Open AI built Chat GPT and others like Google already announced infrastructure of their own. Yet, the mass of founders are coming into the space to build the application layer, not the infrastructure layer. To reference the diagram above, we’re witnessing the exponential growth in apps because in AI, apps are easier than infrastructure.
Web 3 however is exactly the opposite. For years, crypto founders have built the pipes and infrastructure for a decentralized economy. New chains, new protocols, new smart contracts. All of this infrastructure, but they still struggle to find an application that makes it useful. The hard problem to solve in crypto is finding a use case and building a delightful experience on top of the infrastructure. In this case, the application layer is actually harder.
(Tokens represent an interesting blend between infrastructure and application. Depending on the use case, tokens can either be the infrastructure that power a new economy or they can be the application itself, as in the case of Bitcoin. There is also a lot to be written about why the application layer in web 3 is so challenging, but both are subjects for a separate post).
This isn’t to say there won’t be plenty of builders approaching the harder problems — NFTs were the perfect example of this in crypto. Each NFT project tried to create an application on top of existing blockchains — something people could use. This was their attempt at building the harder, crypto application layer. But most of them failed because of how hard this problem is in crypto. It’s easier to build an incrementally better blockchain than it is to figure out what purpose the existing ones serve.
In the world of generative AI, the application layer is starting to look a lot like vertical SaaS. Vertical SaaS refers to software applications that address the problems of a specific industry. They largely use the same infrastructure, leveraging payment APIs, cloud providers, data platforms — but they innovate on the experience to solve their vertical’s problems. We’re already seeing that with chat GPT with AI doctors, resume writers and personal assistants. These can be great businesses but they’re based on the same infrastructure and differentiate themselves with a tailored experience for their vertical. There will be many of these, some very successful, but all highly competitive.
When thinking about hard problems versus less-hard problems, the masses will always flock to the less-hard problems. Yet the harder problems tend to be more impactful once they’re actually solved. Mike Solana takes it a step further in his piece Choose Good Quests. He effectively claims that highly skilled founders and operators have a duty to solve hard problems that help progress humanity forward. Don’t build another SaaS app, cure cancer.
Great companies will be built solving both the application and infrastructure layers in web 3 and AI. But most founders will lean toward solving the easier problems, while great founders should strive to solve the harder, more impactful problems. They create more defensible businesses and usually have a greater impact on the world — that’s part of what makes hard startups such an appealing investment thesis.
Thanks to Kayla, Josh, Avi, Or and Or (yes, there were two) for their feedback!
