DS Dheeraj Sharma Dheeraj Sharma I help non-technical founders and creators build AI automation systems that grow revenue without growing workload. 20 years of enterprise engineering, now in pl...

“The con of vibe coding is that everything stays in the chat or the code that it becomes. And the moment you close it, you have the final outcome as the thing that you have to rework upon. That's the disadvantage of vibe coding. But then came spec-driven development.”

The AI Maker · AI Builders · November 2026

“The con of vibe coding is that everything stays in the chat or the code that it becomes. And the moment you close it, you have the final outcome as the thing that you have to rework upon. That's the disadvantage of vibe coding. But then came spec-driven development.” — Dheeraj Sharma, The AI Maker

Context

Dheeraj Sharma explains why a single prompt to cut dead air and retakes from a raw video only goes so far. He moves instead to a pipeline where each step leaves a file behind that can be inspected and rerun.

Transcript

The AI Maker
Dheeraj Sharma

yet. Still, no word?

Speaker 1

No. No.

Dheeraj Sharma

Okay. All right. So basically, what is happening over here is this whole clip is if I play, so you can see that I have paused over here. I'm not speaking, right? I kind of restarted over here again on the same sentence. So sometimes I fumble a lot. I take breaks while speaking. So if there is like a whole course that I have to record, it would take about a couple of hours to record me. And there will be multiple retakes. There will be silences. There will be, I will say, cuss words as well in the middle when I get frustrated. I'm not able to speak to the camera. It's not easy when you are sitting alone and speaking. Maybe it is a little easier to speak on a live, but when you speak on a camera, at least to me, it's like a little difficult speaking alone. So, I mean, I get a lot of, I would have like a raw file of about, let's suppose, 60 minutes that would cut down to about. 40 minutes of my final video, right? So, 20-25 minutes is usually the waste or fumbles or retakes that are there. So, sorry, I could not like I'll share all this on a file, but…

Speaker 1

there is one question for like a voice enhancing. So when you talk like, do you use any sort of tools to enhance your voice quality?

Dheeraj Sharma

No, I don't. I don't use anything to enhance my voice quality. I have like a mic. Which is like Blue Eti. I use Blue Eti as a microphone to record. So that comes generally okay if I'm like in a silent room.

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.

More from The AI Maker