TLDR

  • Vibe Coding is a new way of coding that uses AI to help you code. It’s suppose to be a tool that helps you code faster and better.
  • Goal: Separate out the hype and see GenAI coding for what it is.
  • Method: Use AI Agents to generate code and see how it works. Use knowledge stores to track the coding and execute the code.

Coding in 2025

Since the release of ChatGPT in 2022, there has been doom and gloom for the tech industry workers. For the first time our jobs might become redundant. It’s 2025 and many more companies has come along and made bold claims of removing the need for human coders but we are still here. It seems there is a lot more hype to remove the depency on human coders than actually seeing benefits from AI coding.

So far it seems that regarless of the GenAI model and the AI agents a human is still required to review the code and make sure its doing what its suppose to. GenAI code will mostly work but it still might get the requirement wrong or more relsitically the requirements given to the AI agent might not be clear. The way a human who might have been around in the company with tribal knowledge of the project would deconstruct the requirements v/s an AI agent with no context of the company will be night and day.

So it seems that AI agents still need some supervision and hand holding, this remains true for now and what is even more apparent is that AI coding agents can greatly boost productivity of human coders. Different camps have emerged on the net, some claiming that AI agents are nice to have versus others claiming that AI agents are the future of coding.

This is core of this blog section, seperating out the subset of the GenAI hype for AI coding agents, figuring out how to use them in my personal and professional works flows.

Why another GenAI blog on AI agentic coding?

So far all the posts and articles are from the perpective of a software engineer which in current market is mainly web development, some are even from the gaming background. Yet I have not seen making claims on AI coding agents from the perspective of a data engineer.

This is also a hard challenge for AI agents as data engineering involves more moving parts. Along with most of the work being done in the cloud, a lot of SaaS products are required and the scale of data is not easily tested locally. So being objective and not biased is hard.

As such I wanted to provide some insights on how to use these new GenAI Agentic tools in a data engineering workflows.

Closing Thoughts

CLARIFICATION: In this series we are not talking about using ChatGPT or Gemini as a coding assistant, we are talking about using AI agents to generate code and run it. This is a very different use case and the results are not the same.

I have been skectic of the GenAI Hype and fear the AI slop, but a lot of claims are being made for Agentic Coding and the FOMO is hitting hard. At moments like this it feels like if you are not at least giving the tech a best shot you are at risk of being left behind and losing out on a critical skill set.

What if AI coding agents are actually really good and I’m not doing it correctly and writting it off as hype and bad tech?

This is what the worst case scenario is, and its worth a shot to go outside the comfort zone and test out the “new” way of working in the even evolving tech industry.