Just like faster computers let us to get away with bloated programs as they are easier to code, LLM will let us get away with more complicated API and sink into technical debt for longer.

Good design is hard, require XP and time.

But ChatGPT creates a buffer against that.

For a while, I'm expecting that libs will become more complicated, not less, and docs not better, but worse, because water takes the path of least resistance.

And just like today we see tools getting more popular because they are faster than the competition, easy API and good docs/tutorials will eventually be a competitive advantage.

Because you still have to understand whatever your AI is advising you to do, and integrate that in your system.

At first, people will, of course, attempt the same things they did with all silver bullets: use it for everything and hope for a miracle.

15 years ago, offshoring was all the rage.

Why overpay local devs when you can hire a 100 underpaids devs far away? But the price of all the projects that went 100% this road and failed re-balanced the market.

Same with NoSQL.

No schema? Let's put all the data in a pile and fingers crossed. Didn't turn out well. Now MongoDB and Redis are like all the weapons in our arsenal, something we use when it's the right tool for the job.

It will be the same with our generative friends.

First, managers will hire a lot of candidates that do perfectly on the questions that we have a good large language model for. Then they will output the first release of patched up stackoverflow answers and GPT outputs, and it will run.

Until all those projects implode months or years later.

It's a cycle, it has always been like that. It's necessary for sane competition, even.

Except this time the magic wand is way more powerful than usual, and it evolves. It will have a project vision and architecture capabilities eventually, although I expect proprietary solutions to fight on that very point.

Still we will finally be back to integrating the now well field tested wonder boy into our practices. Take what's fantastic, leave what didn't work.

And personally, I will once again increase my prices with clients calling me back to save those projects they though they could get on the cheap.

It's a win-win. We get a new marvelous toy, and if we learn to play with it right, we also get more productive, and more cash. Not to mention is damn fun to use.

At least I prefer to think about this future than the potential dystopian alternatives.