Since OpenAI released its latest large language model (LLM), GPT 3.5, I’ve seen lots of speculation about how it will (or won’t) disrupt various industries. When the OpenAI team rolled out Dall-E 2, it seemed like graphic designers were most at risk. That risk hasn’t really materialized, though, because the image generator turns out to have significant limitations. Where we are beginning to see real disruption, instead, is in the software industry. GPT was created by computer programmers and data scientists, and it turns out that the jobs they best know how to automate are their own. Oops!

GPT has already begun to fundamentally change the programming workflow, but I don’t think it’s going to put coders completely out of work. Instead, it’s going to transform the industry’s business model, turning software development into a volume business. And while programmers will be the first to feel the impacts, the ripples will be felt across every sector of the economy as we all adjust how we consume software, as well. As it gets much less expensive to develop in-house, personalized, and boutique apps, I think we’ll see an erosion of institutional advantages and something of a renaissance for smaller firms that are light on their feet. In this way, the decentralizing promises of blockchain may be fulfilled instead by LLMs from OpenAI.

AI is coming for software, but that’s not necessarily bad

The growing consensus in the Twitterverse is that the main use case for GPT 3.5 is writing computer code. It still requires some oversight from a human programmer, but the model can crank out the basic shell for an app with surprising speed. There are now extensions integrating it into popular interactive development environments (IDEs), including VS Code, Visual Studio, Neovim, and JetBrains. Some of these, like GitHub Copilot, cost money. Others, like CodeGPT, are free to use with an OpenAI API key. Either way, programmers can pay mere pennies to get a massive headstart on their code.

At least at first, this will be a windfall to programmers who quickly integrate GPT into their workflow. They can greatly increase their productivity, and until the market adjusts its pricing expectations, they can continue to get paid very well. It won’t be long, though, before software development prices come down very significantly. At that point, developers will have to make up for lower prices by cranking out a larger volume of projects. The ones who survive the disruption will be those who can make this shift.

A few speculations on how things will shape up for developers:

1. The developers who shift fastest toward GPT-assisted workflows will have an important first-mover advantage.

2. With the barriers to entry greatly reduced, there’s an opportunity for total neophytes with no programming experience to get in the game and start selling code.

3. We’ll likely see firms shift away from employing in-house developers, and toward contracting with outside development firms.

4. Developer training programs will focus on higher-level issues like program architecture, software deployment, and computer science theory, and less on the syntax of code.

We’ll all have a different relationship to software now

As the costs of software development collapse, I think we’ll all feel the effects. If it feels like every company has a smartphone app right now, just wait until any business can develop an app for a few hundred bucks. Anyone will be able to purchase bespoke, in-house, or even personalized software to fit their niche needs.

I think one knock-on effect will be a small-business renaissance and the erosion of institutional advantage. Traditionally, mega-institutions like J. P. Morgan have had access to proprietary in-house software and statistical models that give them a big technological edge. Now imagine if credit unions, regional banks, and retail investor co-ops could suddenly and cheaply build their own in-house software to compete. You might see small, agile competitors boom as the big investment banks’ profit margins shrink. And it won’t just happen in finance, although tech-savvy financial firms will move faster than most.

A few more speculations on what this might mean:

1. Companies will build less software for end-consumers and more for their own internal use.

2. Every company must now become data and software savvy if they want to keep up.

3. Small-sized companies (“small caps”) may be poised to retake stock market leadership after a decade in the doldrums.

4. GPT will be a far more powerful decentralizing tool than the blockchain ever was.

Conclusion

The introduction of GPT 3.5 has already begun to disrupt the software industry, with the potential to ripple out to other sectors of the economy. Programmers will have to adjust to a new business model, but they can stay employed if they’re light on their feet. The cost of software development will be drastically reduced, leading to a small-business renaissance and outsized growth for the most agile firms. We’ll all begin to have a different relationship with software, and companies will need to become data and software savvy to compete. All in all, GPT is set to be a major game-changer for both the software industry and the larger economy, and a key decentralizing and democratizing force.