No more Proof of Work

Proof of work is the notion in cryptocurrencies that in order to participate in the network, a node has to spend resources solving complex computational problems. The key is that it doesn’t matter who is running the node or where they are located, because they broadcast their qualifications through effort.

The market for human talent has similar, albeit fuzzier, signals. Having an Ivy League degree “proves” some sort of prior competence. Same with working at a top company. Same with being part of any network of competent people.

However, people have always had a sense that the above were rigged; that there was more nepotism and less social mobility in the system than it showed. Technology was attractive because it allowed people without traditional forms of social proof to show their talent directly by solving difficult problems with the use of tech.

For a time, this direct demonstration of capability felt more meritocratic. For instance, two software developers could raise a little money from friends and family to build an impressive prototype, and that would be sufficient evidence for them to raise capital from VCs and make it big. But these accomplishments are becoming weaker predictors.

At the beginning of the arc of digital technology, say, Steve Wozniak designing and building the first Apple computers, this was incontrovertible evidence of exceptional skill and vision. But since then, developing technology has become easier. More and more people go to college for computer science. Developer tools and open-source software are amazing. Furthermore, there are highly leverageable platforms like the internet, mobile, etc., and infrastructure for rent in the cloud.

In the 70s, there was probably an undersupply of technology companies because it was too hard. What followed, the Y Combinator era of 2005–2020, might have been the golden age of digital startups. During this period, platforms, infrastructure, capital, and founders came together for an explosion in startup value creation. Crucially, startups from the golden era were still capable of building large monopolies and achieving high gross margins.

Today, with AI, it seems it’s becoming too easy to start a technology company. If it’s the case that it becomes too easy to start a company, then proof of work goes away. Consequently, it becomes harder to determine which founders deserve to be funded, because they all have some product, and even revenue. And they get revenue in part by customers attributing innovation credit to founders when they should pass it through to OpenAI and others.

Because it’s easy, there is rapid fragmentation. Many hustlers that before would have become realtors are now looking to build niche software businesses. But if that’s the case, then at least part of the software ecosystem starts to look like traditional businesses: a trucking company, a restaurant, a retail shop. This shift challenges the traditional venture capital model, as VC capital is looking for companies that will conquer an entire market.

Does this mean determining competence for tech founders requires going back in time, looking at pedigree and such? In part, yes. We need evidence of competence, and since prototypes don’t mean much, other signals become stronger. This might work for people who cut their teeth before LLMs. However, now people are using LLMs to get through college, and to do their fancy company jobs, so even those signals are also weakening.

I think it’s more important than ever to find ways to show hard evidence of competence. Maybe there will be an SAT for adults that everyone takes, or employers will start asking for IQ or psychometric tests. Perhaps there will be new institutions that arise to satisfy this need. Having a portfolio of difficult projects that are hard to commission to an LLM is a good start.

My answer has been to focus on raw intellect. Of course, my perception is subjective, but I’ve interviewed thousands of people and managed hundreds, and I think I’m a decent scout of raw talent. One thing that is interesting is that it’s easier to detect raw capability in a younger person, because sometimes experience can look like talent. I expect a new wave of precocious founders that will somewhat reverse the trend of more experienced serial SaaS founders.

None of this is satisfying. I am a big believer in meritocracy, and anything other than concrete results will be a degree or two of separation from true competence, which means it can be gamed.