For any agile team, speed is everything. In the world of early-stage startups, moving quickly isn't just an advantage; it's a prerequisite for survival.

We've all seen it at larger companies: entire teams of developers dedicated to building the internal infrastructure and tooling that makes everyone else more efficient. But for a small startup, that kind of investment is a luxury we can't afford—and one we miss from our days at big companies. Internal tools have to deliver a return on investment almost immediately.

What if they could? We believe the game has fundamentally changed. With the rise of powerful AI coding assistants, a small, resource-constrained team can now create robust, high-impact internal tools at a speed that was unimaginable just a year ago and barely feasible just a few months ago. This is our story of putting that theory into practice, using Firebase Studio and its integrated Gemini agent to rapidly prototype, refine, and deploy a critical tool for our operations.

The Experiment: From Zero to Prototype in 30 Minutes

The experiment began not with a line of code, but with a carefully crafted prompt. With a clear idea for the tool, the first challenge was strategic: how much direction should I give the AI? Should I provide the grand vision at once, or start with the simple, granular instructions where AI assistants excel? I opted for a focused initial prompt and watched it get to work.

‍

‍

The results surprised me, though it did exactly what I said. A working prototype was on my screen in under thirty minutes from the moment I sat down, with most of that time spent testing the AI's output, not writing boilerplate. The assistant handled the tedious groundwork, proving the concept and unlocking value from day one. This initial success quickly evolved into a new, powerful collaborative workflow. Over the next couple of weeks, I guided the high-level architecture while the AI generated the code, allowing us to add features and refine the user experience at a blistering pace. While it took calendar time, my active development time was a fraction of what it would have been—a crucial efficiency gain, as this was happening alongside my other responsibilities.

‍

‍

The Human Element: Where AI Shines and Where It Falls Short

Could the AI have built the entire application alone? Not quite. Its capabilities have clear boundaries (at least in July 2025). While the assistant excelled at scaffolding and standard features, I had to personally architect complex components involving nuanced algorithms. Critical decisions around security, configuration, and system design also required human oversight. The AI is a powerful force multiplier, but it isn't a replacement for an experienced developer who understands the business context and the technical architecture. It's a collaboration, not an abdication of responsibility. 

Yes, we still have to work. 😤

The Future is Faster Than You Think

This experience has given us a new playbook for building. We now understand where AI excels and where it needs guidance, allowing us to leverage it more effectively for future projects. The pace of improvement in this space is relentless; projects that were impossible for AI a few months ago are now achievable. We are witnessing a paradigm shift in real time. The old ritual of grabbing coffee during a long build is being replaced. Now, I grab coffee after writing a prompt.

Perhaps that is the new build time?