"Vibe" Coding today, just Coding tomorrow! 🚀

The future for most app development will be more akin to "vibe" coding than not.

Hi everybody!

So…vibe coding seems to be all the rage these days! 🔥

If you haven’t heard the phrase vibe coding before, it is a term coined by Andrej Karpathy where he describes writing code based on the "feel" of the problem rather than strict planning:

I'm building a project or webapp, but it's not really coding - I just see stuff, say stuff, run stuff, and copy paste stuff, and it mostly works.

The secret sauce to all of this are AI assistants (backed by large LLMs) that have the power to actually turn our ideas and feels into working-ish apps.

Now, to see what all the hubbub is about, I decided to dive head first into the vibe-coding pool. In the following video, I used Project IDX to build a full-stack app powered by a HTML/CSS/JS frontend and a Go backend…all without writing a single line of code (view directly on YouTube or on Twitter / X):

The only coding I did I was describe what I wanted to do in casual English.

While I did need to have a basic understanding of technical concepts to guide the AI assistant on the next steps in some situations, one could imagine that we’ll see increasingly less human intervention in the future.

Vibe Coding will just become Coding

Why do we extensively plan, prototype, and detail out software engineering projects before we start on them? The reason is that the cost of being wrong is often quite high. This cost is paid in money, time, people hours, opportunity, or a combination of all of them. What if this cost goes essentially down to 0? Would this change how we approach software development?

For a lot of us, the answer is a resounding…

When we can go in with a partially formed idea and have a working prototype in a matter of minutes or hours without requiring anyone but yourself and an AI assistant, the cost for just trying things out and seeing what sticks becomes very low. This completely changes all the equations that go into how software is built.

More importantly:

The AI generated prototype becomes the new PRD and…maybe even the starting point for the final app solution.

With all of this said, we can safely argue that AI-generated apps aren’t great for productionizing more complicated scenarios. The apps could have poorly structured schemas that slow database calls. They may have environment/secrets/keys and related details exposed to the client. They may even have other unwanted side-effects:

For the moment, AI assistants are just that…an assistant. You and I are still responsible for being in the driver’s seat and ensuring the right things happen across all parts of how our app gets built.

In the long term, our AI assistants will become more capable. They will be able to better understand more complex application scenarios and system design them in a way that would match what a more seasoned developer would prescribe as the right architecture. This is why I’m very optimistic that vibe coding is here to stay. It is inevitable.

Till Next Time

As always, thanks for taking the time to read this content. What you've seen here is freshly baked content without added preservatives, artificial intelligence, ads, and algorithm-driven doodads. A huge thank you to all of you who bought my books, became a paid subscriber, watched my videos, and/or interacted with me on the forums.

Your support keeps this whole long-running, family-friendly educational circus going! 😇

Cheers,
Kirupa 🥳