Qurious Bytes: Supreet Tare on How Vibe Coding is Shaping Software's Future

The rise of AI-assisted coding tools is reshaping how software is developed. From speeding up prototyping to enabling non-developers to create applications, the impact is profound.

In this blog, we explore the transformative impact of Vibe and AI-assisted coding on software development, based on an insightful conversation between host Amit Chaudhary (CEO & Founder at Enqurious) and Supreet Tare (Founder at Respondr.ai). The discussion covers how AI is revolutionizing the way developers build applications, practical challenges faced when working with AI-generated code, and best practices for leveraging AI tools effectively. Whether you’re a developer, manager, or tech enthusiast, this blog offers a clear overview of how AI-driven coding is reshaping the software landscape and what it means for the future of development.

Question: Tell us about your journey as a software engineer and entrepreneur, and what led you to experiment with AI coding tools?

I’ve spent over 17 years in backend development, working on large enterprise-grade projects. But a while ago, I became curious about how generative AI could improve our development process. I started experimenting with tools, and while I was skeptical at first, I quickly saw how they could accelerate prototyping and remove a lot of repetitive work. That’s when I realized this isn’t a shortcut or a magic trick — AI is like a very capable assistant. It takes time, effort, and the right mindset to really make it work.

Question: How do you see AI coding tools and generative AI changing software development?

The biggest shift I’ve noticed is how fast you can build an MVP now. What used to take weeks or even months can now be done in days. These tools are great at handling boilerplate code and even front-end design — something I’ve traditionally not been strong at. But it’s important to remember: this isn’t hands-off development. You still have to iterate, prompt, test, and refine. AI doesn’t replace developers — it enhances them.

Question: How do you manage culture and risks when integrating AI-assisted coding into existing codebases?

I always treat AI like a junior developer — smart, but not always context-aware. You have to set guardrails. I make it a habit to include a line in my prompts like: “Do not change 95% of the code unless you’re 100% sure, and ask questions if unsure.” That alone saves me from a lot of trouble. I also check everything into version control frequently. These small practices go a long way in keeping things stable.

Question: What does the team structure look like when using AI-assisted coding? How do juniors fit in?

We’re in a very interesting phase right now. Everyone’s learning — some of our HR and ops folks have started building their own apps! So, instead of a fixed team structure, people are working more independently, experimenting and learning through their own mini-projects. But we still need to share what we learn, which is why we’ve created a Slack channel where everyone can post prompts, results, and lessons. It’s very informal, but it works.

Question: What is your typical workflow for building software with AI assistance?

It all starts with the business problem. Once I know what I want to build, I ask the AI to just create the prototype — not the full app. That keeps things manageable. I write very specific prompts, ask it to stick to one screen or feature, and build from there. I don’t even bother giving it Figma files or design references; I let it take a first shot. Often, the result surprises me in a good way. From there, I tweak colors, layout, animations — all in small steps.

Question: Can you share a real example or prototype that illustrates AI-assisted coding?

Sure. I recently built a prototype for my agency’s marketing website focused on voice and conversational AI. I prompted the AI to generate a modern, responsive landing page showcasing three specific services. I kept the scope small — one page only — and iterated. In under a week, I had an MVP that looked good and was production-ready. Compare that to traditional development, and you’re saving weeks of effort.

Question: How do you share knowledge and best practices about prompt engineering within your team?

We use a dedicated Slack channel where everyone shares their best prompts, feedback, tips — anything that helps others get better outcomes. Right now, it’s unstructured, but it’s already proving super helpful. Just seeing how someone else tweaks a prompt or adds examples can be a big eye-opener. Our plan is to formalize this later, but the informal sharing is a great start.

Question: Among the many AI coding tools, which is your personal favorite and why?

Replit is hands down my favorite. As someone who’s always been a backend developer, I struggled with front-end work. But Replit makes me feel like a full-stack dev. I can generate UI, logic, and even deploy — all in one place. It gives me control and confidence, which is empowering. I used to need a team to build things end-to-end — now I can do it solo if needed.

Question: How do you prevent AI from arbitrarily changing code or design you want to keep stable?

That’s a common issue. I always add that one line I mentioned earlier — the 95% rule. Also, I make sure to be as specific as possible in feedback. If a section isn’t working, I’ll refer to it directly, include images if needed, and explain my expectations clearly. The AI adapts better when it knows exactly what you want — and what you don’t.

Question: How do you manage using multiple AI tools together? Does it always work smoothly?

Yes — and it’s a mixed bag. I often use Claude, GPT, DeepSeek, Manus, and others to help draft structured prompts. But funny enough, when I take those prompts into Replit, they sometimes don’t work well. It’s like each AI has its own personality. You have to understand what each tool is good at, and build your workflow around that.

Question: What advice do you have for those starting with AI coding tools, especially regarding mindset and patience?

Patience. That’s the number one thing. Don’t expect to prompt once and get a finished app. Think of AI as a smart intern — they can do a lot, but only with clear guidance and feedback. Start with small projects, iterate, and be okay with reworking things. The more you engage with the tools, the better your outcomes will be.

3 Key Takeaways:

AI coding tools significantly reduce development time but require a clear, iterative approach—think of AI as a smart assistant needing guidance rather than a magic wand.
Clear communication and version control are crucial to avoid unwanted changes, ensuring AI only modifies what is intended.
Collaborative knowledge sharing, especially prompt engineering best practices, accelerates team learning and improves AI-assisted coding outcomes.

AI-assisted coding is transforming software development by making it more accessible, faster, and iterative. It shifts the traditional team structure and requires developers to adopt new workflows and mindsets. Whether you’re a backend developer, a non-technical team member, or a startup founder, learning how to effectively harness AI coding tools can be a game changer in your projects.

Want to hear the full story? Watch the full episode here to dive deeper into how AI-native coding is reshaping software development with Supreet Tare and Amit Chaudhary.