Last week, I started using a new brokerage service for my investments. It is one of those brokerages that can help produce an audited track record, access to more international exchanges, and other bells and whistles designed for the “pros”. Yet, the UI and system was clunky, some buttons didn't work but produced no error messages, and dashboards were often stuck on the “spinning wheel of death” with no data. When I got frustrated, I would go to my CrossFit gym to blow off some steam. The gym has a website that must’ve been built in the 90s, the coaches write every workout by hand on a whiteboard, and we use laminated cards of different weight percentages to figure out how much we should lift for which workout. There is no app to know how many people will be in a class; you just show up and see what happens.

Why am I telling you all this?

Well, as I fiddle with my janky brokerage service and struggle to remember what is 65% of my 1-rep back squat, there is a raging debate on Wall Street about whether code generation AI products – GitHub’s Copilot, GitLab’s Code Suggestions, Replit’s Ghostwriter, Amazon’s CodeWhisperer, Google’s Codey, and more – will create more developers or destroy their jobs. Many developers themselves, especially junior ones, are wondering the same thing.

I propose a simple lens to settle this debate: look around you and count how many products or services you use regularly that you wish are better digitized or “software-ized.” Unless and until that number is 0, there will be more developers, not less.

Judging from my illustration of my personal life, you can safely surmise that my answer is decidedly not 0. Not even close. And I live in the United States and work in tech. If you live outside one of the more technologically-advanced or Internet-connected countries, that count will likely be higher. There are way more technologies that ought to be built that aren’t. AI-assisted programming only enables more people to fill that void.

Despite the zero-sum nature of Wall Street, if you view the world from my proposed lens, it’s quite obvious that AI is an accelerator, not a detractor, of unabated global developer growth.

Does Open or Close Source Matter?

An orthogonal debate is whether open source or closed source AI is the right approach. This debate was raised to a new level of attention recently by this leaked memo from a Google AI researcher published on SemiAnalysis. The researcher argues that neither Google nor OpenAI can maintain a competitive moat in the long run as open source AI alternatives improve rapidly.

I’ve been immersed in the open source industry, as an operator and investor, for almost 10 years. I’ve seen up close that open source tends to produce superior technology and has a built-in set of transparency and norms to build strong trust. Thus, the central argument from this leaked memo is preaching to the choir for me.

But I’m not “religious” about open source. While I generally prefer the open source approach to most things, on the issue of AI abuse, I see valid concerns of a powerful model being completely open with no restrictions being misused by bad actors. There is a long way to go to effectively regulate these concerns.

Closed source solutions also have some competitive advantages, dare I say moat, if they are packaged well and have superior developer experience. That’s because different types of developers want different things from their tools. There are generally two genres of developers: application developers and infrastructure developers.

Application developers tend to like out of the box solutions, so they can build their applications quicker, because they are motivated more by the real world problems they are trying to solve, not necessarily the tools that will help them solve those problems. So they tend to gravitate towards closed source solutions or at least they don’t have strong opinions about whether a solution is open or closed, as long as they can make use of it easily.

Infrastructure developers tend to gravitate towards open source options, because they like to look under the hood, tweak and tune to their preferences, and exert a higher degree of control over the software or model. This group likes to build tools, not applications, and is motivated by the technology itself. While some are motivated out of pure curiosity, many also do it out of responsibility, because they often have day jobs where they operate and secure cloud infrastructures for large enterprises, so having under the hood access and control is critical.

All in all, there is room and reason for both open source and closed source AI products to exist and shine because they each cater to a different developer audience. Thus, the open vs closed debate doesn’t really matter in the end, because the overall developer pie is growing, AI is pushing that growth further, and new developers will make good use of open or closed solutions depending on what type of developer they are.

Tech Company Creation

While the macro trend of global developer growth won’t reverse due to AI, despite Wall Street’s needless worries, one thing will change: tech company creation.

AI-assisted coding products drastically reduce the barrier to entry to becoming a “developer.” They also make experienced developers more productive, which really means more “creative”, whether it’s solving tough technical problems or solving new real world problems using technology. So more people will have the wherewithal to fix some problem in their lives by building software. Maybe it starts by building a small app to improve a local CrossFit gym. If the app is useful, it could be bootstrapped into a small SaaS product. While all these daily life problems seem small, you never know what the idea could grow into. The company, Mindbody, which IPO’ed in 2015 and was taken private by Vista Equity in 2019, is essentially a scheduling and back office SaaS for gyms.

More digital products will be created, but the companies that build them will on average be smaller. More of them will be bootstrapped, self-funded, generating revenue earlier in their lifecycle, and not funded by VCs. There will be more and more one-person teams building meaningful products and companies. While bootstrapping software has always been a thing, supported by communities like Indie Hackers, AI-enabled programming will unleash the biggest wave of technology creation yet, not for the sake of creating, but because, if we just look around us, there is so much that technology has not yet touched but should.

Midjourney, the AI-powered image generation organization, is the poster child of this trend. Despite its outsized influence and impact, it currently employs 11 full time employees and is self-funded.


Of course, a company like Midjourney would never raise a ton of VC money to need a liquidity event, like an IPO, to make investors happy. Perhaps that’s why Wall Street is worried. More small companies don’t exactly feed the Wall Street business model.

That view is, of course, quite myopic and short-sighted, when we are barely in the first inning of generative AI and there will decidedly be many (many) more developers building in the next decade than the last.

But I’m not in a hurry to correct this misreading of the future. Wall Street’s myopia is my alpha.