I have been writing a few posts about improving your vibe coding efficiency lately and this post is one more such contribution. I am building products since 6+ years and this post is about things that you can just start doing and see your vibe coding skills improve with huge margins.

One thing to start off, this is neither a guru-kind of post where I will claim I know something you all don't know, nor it is that I am any more intelligent than you. ItItss simply just my personal experience.

Here are the things (by increasing order) -

1. Building from 1st Principles : Most vibe coders start off by just chasing something extra complex which might look achievable till 90%, but gets much more complicated in the later 10% due to the engineering limitations of human or artificial intelligence. So, rather than telling the AI rigorously to "fix the code", see the code, send parts of the code to other reasoning models, and think - "Why at the point, this error is occurring?"

If you start thinking in less complex questions, your reasoning capacity becomes strong thus making your overall interaction with vibe coding applications stronger.

2. Think in terms of structure : This is one of the key things vibe coders are missing. Most of us just write a prompt and expect AI to come up with a ready-to-deploy version. But honestly, AI is not our senior developer (yet).

If I am to start with a new project today, I will come up with the list of all stacks I will be using - Db, auth, vectors, ML libraries, etc. and then create a list of these libraries, and a folder structure to start off. I might use ChatGPT to create this but the main point is, I get it pre-built before starting to vibe code. Once done, I will let my IDE take over it.

3. Documenting & Always be Knowing What Your Code Does : One major setback for vibe coders is that most of them don't really understand what part of their code is doing exactly what? That's the reason why a developer can do atleast 3 times better vibe coding than a regular person coz he can read the terminal for errors, see through code and apply his intelligence in writing prompts that will fix things.

Now, don't start learning programming because of this. You are way better than that. But just get knowledge of how backend like HTTP, CD/CI pipelines, SQL, guardrails, evals, Websockets, API, Redis, etc. works. Once you know what this is, you will be much better at mindful thinking and building stuff.

About documenting, you can just create one file (mind.md or something) that will contain all major changes and logs of things you are doing. Even you can go a step beyond and create proper documentation for you, your users and future dev team to help understand what your code is about. You can use SuperDocs[dot]cloud for using AI to create documents without actually writing it.

At last, the advice is to think and have knowledge of stack like a developer but execute with AI. That will help you make much better apps at scale. I know, I might have missed a few things, feel free to add stuffs in the replies.

Also, if you are building this by yourself and got stuck somewhere or have any question, please ask me, I will try to respond you.