When I signed up for Lex, the word processor that incorporates GPT-3 to help writers quickly progress through momentary writer’s block, I imagined my productivity skyrocketing. The very first thing I did with it was to write a post about how generative AI would transform fintech, and I used Lex to help me write it.

Then something unexpected happened. I got the worst case of writer’s block ever. I’m sure no one but me is keeping track, but that’s why I haven’t written a new post in a month. For the past four weeks, I would try to get my mind to think about interesting changes in the world of startups—specifically fintech—and it just goes back to thinking about generative AI. I’m obsessed.

I’m chewing on it, digesting it, metabolizing it like some sort of hallucinogenic drug. I catch myself lost in thought about potential GPT-3 business ideas, and everything that comes to mind is too easy, too obvious, or too indefensible.

When I wrote my first post on generative AI and fintech, something I didn’t mention because it didn’t fit the message of the post, is that I wasn’t concerned that generative AI would have a big impact on fintech because most of the work done in fintech doesn’t require generating something new. We need to connect systems and maybe analyze data, but we’re mostly not creating. Except, I thought at the time, for the code that we write to do this work. But I felt fairly certain that truly useful code generation was a ways off. Then a few weeks went by and truly useful code generation happened.

You’ve probably seen this example and better, but just in case, here’s an example of ChatGPT writing code to generate an animation of rain in a web browser.

@joshsowin ChatGPT makes simulated raindrops using p5.js #processing #ai #aiart #gpt

Tiktok failed to load.

Enable 3rd party cookies or use another browser

Every developer I know is talking about this right now. We are all transfixed. We are playing with it, and we are already using it to write our code. Coding with ChatGPT, in my mind, is far more valuable and powerful than using it to write prose.

The takes I’ve seen are about how AI could replace junior developers, how we could all lose our jobs, and how developing has just turned into managing the output of generative AIs.

That last statement is the closest to the truth. But I’ve been thinking more deeply about what this all means? How will this play out in the market? What will software become?

Then it struck me. I knew. I know.

We are about to enter a time of vastly more code running on processors than has ever run before. The next phase of fintech, social media, word processing, games, commerce, EVERYTHING will be about personalization. This means personalization not just for consumers but for businesses. Over the past 50 years, to create great products, experienced product managers have had to work some magic to know how to find the greatest common UX denominator across a range of customers and business cases to build products that suited the largest possible addressable market.

Now, huge backlogs of customer ideas and one-off requests squirreled away in product management tools will become viable.

As we go through a huge round of tech layoffs, I’ve noticed some hand wringing and wondering out loud what this all portends for the tech industry. The New York Times did an article about the shrinking job market for recent grads, and an incorrect interpretation of this story is that there is less demand for software.

Au contraire. The demand is there. Customers want changes, updates, features, capabilities. But the economics of building new software—especially in big tech companies, where developers were beginning to feel entitled to total compensation packages worth as much as a $1M per year, made the economics of adding certain types of features difficult. The math is simple, right? If it takes a person year to code it and maintain it, then it better generate well over $1M per year in revenue.

If the same amount of code could be built in a month or a week, then the project would get a green light with hardly any balance sheet scrutiny.

Here’s a personal example: At Kelsus, one of our secret advantages for hiring great people in Argentina is that we have a flexible work and payment structure that allows developers to control how they are paid at the same time as enjoying the benefits and protections of employment by a local company. Over several years, we’ve built sophisticated business processes to manage all this, and I think of it as our mini-fintech. There have been times when we’ve talked about moving away from spreadsheets, Slack, and email to custom coded systems—after all, we are developers. But every time we’ve looked at it, we’ve decided that the amount of code we’d need to write isn’t worth it for our small team. That time and energy would be better spent on a different part of the business. With AI code generation, the equation changes.

In this new world of vastly reduced price-per-feature, each customer—whether in-house or external—will be able to buy products do exactly what they want. Software companies that embrace this new paradigm efficiently by building systems and architectures that enable it will be the most successful.

This new world is here today. Developers are already bragging to each other, teaching each other, and accelerating their craft to unheard of levels of output. 10X developers, 100X developers, and 1000x developers are all going to become real. The future’s 1000X developers will still need to reason, think, combine, derive, and plan. They just won’t need to toil as much. I also still think they’ll be paid very well because they’ll be responsible for building vast amounts of business value. We’ll set aside concerns about whether all this value will be distributed equitably for another post.

The question to me now is whether this all happens over the next 5 years, 5 months, or 5 weeks!

—Jon Christensen

Startup Win

What is the endgame with software development…