God of the gaps and the future of programming

gisbi@homepage ~<br>← blog God of the gaps and the future of programming<br>2026-05-19<br>◉ 15 ♥ 2 [share]<br>Being a programmer nowadays is a cognitively challenging job. Or at least, it has been for me.

For the better part of one year, I've been asking myself if I'm doing anything more than effectively cosplaying as a developer, and I've been thinking about the God of the gaps often.

The reality of the situation is: in a three years time frame, AI coding agents have evolved from situational helpers that could help you debug small issues or teach you a few tricks you didn't know about, to fully fledged beasts, capable of handling entire projects from start to finish.

Code quality and coping

A common argument I see being thrown around a lot is that vibe coding seems to work on the surface, while actually producing more bugs and hallucinations than working features.

This argument was legitimate until mid-2025. The quality of the code generated would exponentially decrease with the feature's scope and complexity.

However, I still see it being used today, which is, at best, willful ignorance or, at worst, a complete lack of effort from the developer's part.

Skills and sub-agent driven development, with comprehensive planning before writing a single line of code has been, at least for me, the straw that broke the camel's back.<br>The number of times I've needed to manually correct code with this implementation strategy has decreased so much to become basically a rounding error.

One could argue that the quality and complexity required at my current job to not need any manual coding is quite low, to which I would agree with, but I've yet to talk to any developer, writing much more complex code than me, who isn't in a similar position.

Developer of the gaps

I want to avoid controversial statements. So, the safest way I can phrase this is: I think we can all agree that in three short years we're at a point where writing manual code is becoming increasingly rarer each day.

I see many developers going through the first of five stages of grief, with the first being denial . They dismiss the tool, because embracing it would force them to face the fact that it's doing a better job at writing the code, while also taking a fraction of the time it would've taken them to write it.

Like the God of the gaps, the "Developer of the gaps" squeezes themselves inside an ever-shrinking gap that AI can't fill yet.

First, it was accessing the codebase.<br>Then, it became holding enough context to not start hallucinating.<br>Now, it's handling things outside the repo itself, even though that seems to be going away soon too.

I'm avoiding a blanket statement because I'm sure people writing mission critical code are not affected by this, but the run-of-the-mill dev working in JS (group of which I'm a part of) is.

Cognitive dissonance

I actually enjoyed the manual part of writing code. I like typing, so getting my hands dirty was one of my favorite parts. Not only that, but no matter how well you understand what's being written by the AI, there's no substitute for thinking through and typing the code yourself.

So..how do we fix this?

I've seen people suggest sticking to vibe coding for work and manually writing your personal projects, so I tried that (I've always been into game graphics, so I've been working on a game engine in Go called "Ristretto", WIP).

What transpired so far is a deep sense of unease and uncertainty during the whole process.

There isn't a second when I'm working on Ristretto where I don't think "this could be done in a day".

Obviously, I'm aware that the trade-off of vibe coding it would completely invalidate the reason I started working on it in the first place, which is learning how game engines work under the hood.

Still, I can't shake off the feeling that if, and I apologize for repeating myself, after only three years of vibe coding I'm already struggling with these thoughts, assuming progress doesn't just grind to a halt, then it's just a matter of a couple more until coding manually will be an activity done purely for fun and nothing else.

Conclusions

There's effectively two possible roads going forward.

If this is a bubble, when it eventually pops the real price of using AI hits consumers, and vibe coding becomes a luxury item a few people/companies can afford to use daily.

If, for some miracle it's not a bubble and this is our new reality, then it's quite possible a new paradigm is created.<br>We've seen time and time again that whenever humans can automate work, instead of working less, they just produce more with the new standard.

It's not crazy to think that a new symbiotic, instantaneous "coding" language that uses AI under the hood, becomes the new way to write things. A good friend of mine actually thought about this already, so if you're interested in the idea, I highly recommend his article.

Thank you for reading :)

$ type "cd blog" to go...