- Cognition
- Posts
- Marketer's notes on vibe coding
Marketer's notes on vibe coding
We got the vibe horribly wrong

Imagine having the strength to move mountains, but you use it to lift furniture. Marketers are doing the same with AI. Here’s why.
Marketers are still playing around.
Our mindset is, at most, that of a child gifted with a motor-controlled car.
We are just playing, not solving. It’s because we are in the exploration phase, and in awe of AI.
“Oh, I can train a GPT to talk like me”, “Oh, I can create a time tracker”, etc. Remember the time Notion templates were a thing? We are doing the same now.
Our perspective is not, “How will AI solve my problems?” Yet.
There is good and not-so-good side to it.
Good is exposure. There is a popular product concept, “Everyone has one habit tracker in them.” As I understand it, it means we can build basic apps to know if we enjoy product-building. If this is the case and you are aware, great! Keep vibe coding and shipping.
The downside is we are distracted. We are marketers. We deal with multiple flows, channels, and apps daily. This is a personal opinion, but shouldn’t we be making our lives easier first? Things like automating repeated tasks, data analysis, creating workflows to reduce manual work, content creation, etc. Instead of building habit/time trackers or chatbots.[1]
Aren’t we spending more time than needed on building tools, while we can use the same AI to directly save time and improve efficiency?
I am not saying we shouldn’t build products. I will probably write down five more web app ideas by the time this essay is published.
Vibe coding trended for good! It generated a curiosity in non-technical people to ship their ideas fast. It shifted us from limited, no-code platforms to a more powerful approach. We will witness more non-technical people build and hit sexy products and MRR soon.
But I’d rather solve my problems first before I solve for the world.
Vibe shouldn’t be the limelight
Vibe coding is a tool to get where we want to be faster, not the limelight itself.
We are excited by “I used AI to build this.” Instead of the product itself: “X product solves Y. AI makes it possible and more efficient for me.”
The excitement is beautiful as we discovered a new skill, but we shouldn’t move away from the fundamentals of product building - “make something people want.” With or without AI, users don’t care how it is built but what it solves for them.
We don’t always need apps
I noticed our instinct as marketers is to build web apps. Prompt the problem, get the code, deploy, and now we have a functioning website.
I was this person who made websites for everything, and speaking with more vibe coders/marketers made me realize I’m not the only one.
Now I check for three things before building a web app:
Is there a solution already that I can use directly?
Can I connect two or more apps to solve my problem?
Can I create a project in Claude and contextualise it?
My favourite tweet
”Once everyone can build apps, we will know it’s not about building apps” has to be the best tweet I came across during this trend.
We underestimate how complex coding is and generalize it to prompt = web app. Vibe coding is fun and games until the output isn’t what you expect, and you have no clue how to deal with it.
It’s hard to be a builder without understanding the product in and out. I’d be paranoid if I didn’t know how my product works. Of course, that doesn’t mean I learn to code. That defeats the purpose of AI. But you need to know what file contains what. What to do when a fuck up happens. The zero-to-one process, etc.
If I can give newslettercasestudies.com’s example, I prompted AI to give me HTML and CSS files separately so organization becomes easier. Plus HTML because easy to understand. I don’t know the function of all HTML tags, but I know what each tag on my blog performs.
You don’t need to know the skill. But you need to know your product. It’s non-negotiable.
Solving for myself
I am not saying this is how marketers must approach vibe coding. It’s not my place. I am only sharing how I work around AI.
So far, I invested my time plus saw my friends build products they don’t want. Or solve a problem that didn’t exist. It was good as an “exploration phase” and now we move on. We move on to solving problems for ourselves - where we can create workflows that eliminate hours of manual work.
In my case, it could be as simple as setting up Spidey 🕷️ in our marketing community to check up on our members:
![]() Spidey solves simple, automated messages and engages the community | ![]() What I plan to do with Spidey |
Or as complex as some ideas I am building on currently:

Not fair to give you gyan if I am not walking the talk
It stems from a philosophy of personal software[2]. I love the idea that you can make an individual product used by you and only you for your custom requirements. No users, no monetisation, you have control over the data, and it does the job.
Learn to learn tools
This tweet sums it up, so I won’t take any more of your time:
ai this ai that; i say learn to learn tools fast. the most premium tools today are the worst they'll ever be and we'll only see more advanced versions or new tools.
only way to stay ahead of the curve is to know how to navigate through tools and apply it effectively asap.
— Vikra Vardhan (@vikravardhan)
1:52 AM • Apr 12, 2025
Footnotes
1/ I noticed many marketers build products outside their niche. It’s not a bad thing, but I feel it lacks authority. Some marketers I know built products in Finance and Health. The products are good, ngl, but it’s hard for a new user to see why they should use your product over industry leaders. Just a half-cooked thought. Happy to hear your views.
2/ This is one of my favourite blogs; it perfectly sums up personal software. If you want to read more, there’s this too.