The case for vibe coding

Before I start, I must confess that I detest Vibe Coding. If not the movement itself, then certainly the principle behind it — which flies smack in the face of discipline and discernment built over time.

A lot of the poster children are indie hackers leveraging their popularity to push easier ways to make money, and product managers eager to imbibe the “fire your dev” philosophy only to faceplant at the first pitfall.

The knee-jerk reaction comes from an element of threat of dev jobs being overthrown — a threat that has endlessly been harped on since the explosion of ChatGPT in 2022. But much of the gripe also stems from the celebration of “just about” functional code paraded as if it’s solving the world’s problems. When in reality, it often feels like a precocious junior dev got bumped up to CTO because they solved one production bug in record time.

But as stricken with rage & confusion I might be with this “trend”, there have been several critics who’ve articulated the dangers and the problems of Vibe coding quite well. The internet is now flooded with them — and for good reason, considering the looming security threats and the potential for on-call pager duty nightmares, all because some product visionary wanted to ”feel the vibes“.

But I wanted to challenge this norm as a mental exercise. To be that reluctant advocate for the devil.

So, I’ve been trying really hard to think from an AI optimist’s viewpoint. An uphill battle, considering most of my feed is peppered with AI doom prophecies and Terminator-esque scenarios.

There is one thing that AI-assisted/augmented coding clearly helps with: it democratizes the creation of software. The fact that someone with minimal or no coding knowledge can create a bespoke piece of software opens up a new paradigm for how we think about custom apps.

Generally, the apps we build must cater to n number of users. Regardless of how generic or specific the feature set is, we usually aim for a general solution. But when n = 1, you can tailor the software exactly to your needs.

This opens up a realm of possibilities where the power shifts to the consumer — instead of waiting on some SaaS company whose roadmap is two months deep into assessing their own term sheet.

A not-so-distant extrapolation would be the creation of custom apps through a conversational UI. For example, prompting: “Build me an app that consumes sleep data from my smartwatch to recommend optimal alarm times based on my circadian rhythm.” And BAM — you have something to get started with.

This paradigm shift, while disconcerting, may not eliminate dev jobs in product creation. Instead, it could relocate responsibilities to other areas and create new roles — especially in standardization and data protocols. After all, this bespoke software era could usher in a nightmare of unwieldy data structures.

Of course, most of these apps will be created with only the ”happy path” in mind. Problems will emerge — and they’ll piss off a lot of people, especially seasoned devs, who’ll see their values and principles tossed aside.

Yes, I’m aware I may be using “Vibe Coding” and “AI-assisted coding” somewhat interchangeably. But my broader point is about the ubiquity of creation this might lead to.

I’ve really wrestled with the idea of what fresh hell this lowering of the bar for software creation could unleash — until I watched this video:

The beauty of that video doesn’t lie in the software’s elegance. It might be riddled with bugs, and there are probably better ways to handle the UI. But the fact that someone, with a code editor and an AI, managed to build something to improve someone’s life — that’s a slam dunk for the ethos of software creation.

I can only hope for a more inclusive world of software building — one focused on problem-solving, not gatekeeping goblins hell bent on proving how html is not a programming language and why PHP sucks.

Now I’m in no way asking everyone to whip out their AI code editors and churn out software after software with obvious glaring security flaws. Vibe-coded projects still belong in personal contexts — weekend experiments, or projects like the one in the video.

It should not touch anything marked production even with a barge pole. And we’re long way away from complete AI projects being production ready. There’s a greater chance of Trump winning the Nobel Peace Prize. The odds of which are now based on dice rolls.