Is the College Student Startup Pipe Dream Dead?
The college startup pipe dream
I write this article to pose a question: Will the next trillion dollar startup be started by a college-aged kid in his dorm?
Can the pattern of Microsoft, Meta, Amazon, Apple, and Google be repeated? (Bezos being a sort-of exception to this, as he was ancient when he started Amazon at 30).
An analysis of historical successes
What’s interesting is that all these incumbents started as consumer startups, and most hyperscalable startups that have started in the interim are B2B SaaS. Looking at the exceptions, even the newest fastest growing social media incumbent (TikTok) was started by a company not an individual. Past that, the next biggest consumer innovation was ChatGPT, and that was created by a team of research scientists, a far cry away from a guy and his friend in a dorm.
Then there are the Uber’s, the Doordash’s, the AirBnb’s, et al. But these companies are quite limited in their scope, and clearly nowhere near trillion dollar scale. It doesn’t seem like they have a way to get there either. I don’t say this to downplay their extremely impressive achievements, just in pursuit of an accurate understanding of the environment. Further, these founders seemed to be in their late 20s/early 30s when starting their business. Even the hard data suggests that most unicorn founders are in there mid 30s.[1]
I think there are a few reasons for this. The first being that these big companies have become effective monopolies on scales never seen by capitalism before.[2] The second being a relative stagnation in the technological platforms being used. People are sort of set on phones, laptops, and the web. They have been out so long that almost everything within the bounds of cognition has been tried.
Apple and Microsoft took advantage of the rise in personal computers (and then Apple, again, took advantage of the rise of smartphones). Amazon, Meta, and Google, all were built off the rise of the web, and then had a resurgence with smartphones. On the backend, there is the switch to cloud and horizontal scalability, which a number of these companies also took advantage of.
The ways people interact with technology have only gotten sub-linearly better since then. We have been treading in the bottom of the log-loss function for so long, all the easy gains have been optimized away.
Looking towards the future (AI)
The obvious next question is about AI. Do the improvements in AI constitute a major platform change? If so, then the college student startup pipe dream has a glimmer of hope.
I generally think so. This is for a few reasons.
The first reason is that this technology is not really an incremental improvement over the previous RoBERTa type models, and I think few people understand how surprising it is that these scaled markov chains approximate intelligence.
The second reason being that ChatGPT and other LLM based tools I have hacked on/played with are much better to use and interact with than their alternatives. For example, one of my friends is building an AI-first CRM for Real Estate agents, and customers seem to respond well to it. Further, my intuition tells me there are a lot of unknown unknowns when it comes to the application of this technology.
The last reason is that I am sort of blithely optimistic that technology can be growing on a Moore’s law-esque scale for a little bit longer and that someone like me (a random college student) can actually change the world on the largest scale. Call me foolish.
I think looking at OpenAI and Anthropic, you can be fooled into thinking that no one can compete with them, I think you would be wrong.
I generally think ChatGPT as a product is of limited use to anyone who is not a programmer or student. Maybe writers can game some feedback and overcome writer’s block, but that use case also seems extremely limited. Even in those use cases, the UX is kind of just bad.
There are a lot of ways to package and build around LLMs. I think a clever college student will put some puzzle pieces together in such a compelling way that it will change the way we think. I can’t really imagine what that would look like (otherwise I would just build it), but this is what my intuition tells.
Apart from raw text and on the model side, it’s also surprising how model performance scales with data. I think pretrained transformers can be applied to areas other than raw text, and that encoding and decoding higher order relationships is extremely powerful tool. If you are a curious person, I would highly recommend checking out nanoGPT and learning about making your own transformers. Do not think that only OpenAI or Anthropic can produce model level innovations. There is a large market for “foundation models” as the VCs call them.
Postscript
Credit to my extremely smart dad who’s conversation inspired this article. He was trying to get me to focus a little more on college instead of writing useless blog posts or playing with the shiniest new technology or idea.
tl;dr: I don’t think so, but some would call me naive.
Footnotes
[1] https://finance.yahoo.com/news/typical-unicorn-founder-started-business-120026888.html
[2] I say “monopoly” in the same way Peter Thiel says monopoly