Guide to AI-Driven Development

The way we used to build software has changed forever.

After a long weekend of vibing and running background agents for production grade code. I have come to some reckonings about the future of software development. This realization has also led me to principle with which we can build software in general.

The core philosophy of AI-Driven Development is that a developer is now moving from being a builder to an operator.

Each software class has its own unique characteristics and requirements. Understanding these classes is crucial for effective AI-driven development.

Category	Examples
UI-Oriented Software	Web Applications, Mobile Applications, Desktop Applications
Backend-Oriented Software	APIs and Microservices, Data Processing Systems, Automation Tools
Embedded Systems	IoT Devices, Firmware Development, Real-Time Systems
Enterprise Software	ERP Systems, CRM Systems, Supply Chain Management

In early 2023, I was skeptical about the AI hype. I believed that AI was overhyped and that it would not have a significant impact on the industry.

As the model improves, agents get better, and the ecosystem matures. I think the debate of build vs buy, ToC starts to shift.

You now have the ability to build internal tools for every specific problem of the organization in a verticalized manner. With opus, sonnet, gpt 5.2, gemini 3, you can do it to a point that you don't even need a developer to build these internal solution, a product manager or an analyst might suffice.

The cost of building these tools will be lower than buying a SaaS solutions, and the time to market will be faster.

The control, customization, and integration with existing systems is going to be better.

So, in 2026, I believe NIH syndrome is going to be on steroids.

You can start with a blank slate, or you can seed the project with existing code snippets, libraries, or frameworks.

Category	Examples
UI-Oriented Software	Web Applications, Mobile Applications, Desktop Applications
Backend-Oriented Software	APIs and Microservices, Data Processing Systems, Automation Tools
Embedded Systems	IoT Devices, Firmware Development, Real-Time Systems
Enterprise Software	ERP Systems, CRM Systems, Supply Chain Management

In early 2023, I was skeptical about the AI hype. I believed that AI was overhyped and that it would not have a significant impact on the industry.

As the model improves, agents get better, and the ecosystem matures. I think the debate of build vs buy, ToC starts to shift.

You now have the ability to build internal tools for every specific problem of the organization in a verticalized manner. With opus, sonnet, gpt 5.2, gemini 3, you can do it to a point that you don't even need a developer to build these internal solution, a product manager or an analyst might suffice.

The cost of building these tools will be lower than buying a SaaS solutions, and the time to market will be faster.

The control, customization, and integration with existing systems is going to be better.

So, in 2026, I believe NIH syndrome is going to be on steroids.