Vibe Marketing

His words: "we said, why do it, bhai."

He also built a fully automated end-to-end AI content workflow that produced output at fourteen or fifteen dollars a go. It worked. He killed it too, on arithmetic that only makes sense in India: a person at three thousand rupees a month will make the video. He is still running a business news operation that produces at under a hundred rupees an item — and he will tell you the audience does not enjoy it.

That is what this course is, and it is not what AI courses usually are.

The premise is narrow on purpose. Vibe marketing is his own term, taken from vibe coding, and he defines it as a method of execution rather than a new discipline. No new channels. No new theory. The marketing you already do, with AI doing the doing. He says on camera that the course is for people who already know marketing, and he means it — ads, funnels, SEO and analytics vocabulary are assumed from the first lesson.

The whole prompt method is three words. Write whatever you were going to write. Then append "Refine this prompt." Then run the refined one. That is it, and it is better than most of what gets sold as prompt engineering.

Six cases, each run live with real inputs and real outputs.

Google Ads scripts written in plain English — including four production rules he endorses: click-through under 1% for thirty days, pause it; over ₹10,000 a day, alert me; more than ten conversions, raise budget 20%; fewer than two, label it underperforming. He notes that scripts like these used to sell for five hundred to two thousand dollars on marketplaces he can no longer find.

Market research run through three deep-research tools simultaneously on one prompt, so the comparison stays clean — and a verdict on each. Gemini when you want to study something. ChatGPT when you want prose you can publish. Grok when you want human opinion, because it is Twitter's. One run read 109 websites in five minutes, against the ten to twelve days he used to spend researching a single video manually. And the moment that matters: one of the tools inserted an image that was plainly fake, which is exactly where a human has to be standing.

A thumbnail pipeline from a 49-minute podcast, with the rule stated three times — research, then idea, then image. Never straight to "make me an image". A product photoshoot built from a single PNG, including a back pose generated by telling the model in advance that the shirt is plain from behind. And a year of YouTube analytics uploaded as a CSV and turned, in just over a minute, into work he estimates would take a person two days.

Then the cost discipline, which is the part nobody teaches. Run the cost test before you automate anything. Know that an over-powered model on a simple task can cost thirty to forty times what the right one costs. And know which work AI actually eliminates — his framing is the labourer and the mason, and it is more useful than the usual hand-waving about jobs.

Three warnings, plainly.

This is the most date-sensitive course in the catalogue. It was recorded around the middle of 2025, and roughly a third of its concrete content is a snapshot of a tool's interface or price. The model-comparison lesson is already wrong — it discusses a lineup that has since been replaced entirely. The durable half is the reasoning: consumer change, cost discipline, how to handle bias and hallucination, the funnel logic, and when a human has to intervene.

The bonus module is a full workshop on building custom GPTs. He says elsewhere in the same course that he stopped using custom GPTs about six weeks before recording, because ChatGPT gained long-term memory across chats. Both things are in the product. Watch the workshop for the method — the instruction template, the story taxonomy for training it on how someone thinks — and not as current tooling advice.

And Module 5 is five downloadable resources with no video attached.

Taught in Hindi.