I Built a Windows App with Google Search AI to See if Free AI Can Actually Code

The Google Search we used to know is officially gone. Nowadays, whenever you search for something, you’re greeted with an AI summary or the option to enter "AI Mode." This essentially gives you a free, unlimited AI chatbot sitting right inside your search engine.

That got me thinking: Could I actually build a fully functional, lightweight Windows application from scratch using nothing but free Google Search AI?

There are plenty of incredibly powerful, dedicated AI coding agents out there like Claude, Cursor, or Devin. But let's be real—they cost money. Their free tiers run out quickly, often before you can even finish a basic app. Google Search, on the other hand, is completely free.

To make things more interesting, I decided to do this the old-school way. No fancy IDEs—just me, Google Search, and good old Notepad. Here is how the challenge went down, the obstacles I faced, and the ultimate verdict!

The Target: Solving a Legacy Windows Bug

For this challenge, I needed a real, practical app to build. I decided to tackle one of the most annoying, legacy issues on Windows: webcams resetting to default settings after every PC restart.

Windows handles webcam hardware configurations using legacy DirectShow APIs. These APIs don't natively cache your presets when your computer power cycles. I wanted an app where I could adjust my camera, save those custom settings to a file, and restore them with a single click.

To set up the initial file structures and save myself a bit of manual setup, I also deployed a small, offline local AI agent (Gemma) to assist with file scaffolding, which worked perfectly in tandem with Google's search results.

The Reality of Search AI: Hands-on Manual Labor

Building an app with standard Search AI is a completely different beast compared to using premium coding agents. Premium agents do everything for you—they create the directory, write the files, and run the code.

With Google Search, you have to do the manual labor. You are the one creating files, copy-pasting code, and manually triggering the compiler in the terminal.

The Obstacles & Debugging Loop

Almost immediately, I ran into classic development hurdles:

The Classic Build Fail: When I ran dotnet run for the first time, the compiler threw multiple syntax and missing reference errors.
The "Code Cut-Off" Limit: Because the front-end C# code (Form1.cs) was so massive, the Google Search chat window kept cutting the code block off mid-sentence.
- The Fix: I took a quick screenshot of the formatting issue and sent it directly to the Search AI. It immediately understood the issue, apologized, and smartly split the massive codebase into separate, manageable files: Program.cs, ConfigService.cs, WebcamController.cs, and Form1.cs.
The Silent Crash: At one point, the build succeeded, but the app would silently crash on launch. The cursor would spin for a second, and then nothing. I fed the crash behavior back into the search bar, and the AI correctly identified a project configuration issue that was easily fixed via Notepad.

Fixing the Infamous "Cross-Talk" Bug

Once the app finally launched and successfully recognized my Logitech C920, I started testing the sliders. The brightness and contrast controls worked great, but adjusting the anti-flicker slider unexpectedly started changing my sharpness slider too!

I asked the AI why my hardware controls were bleeding into each other. It diagnosed it as a "variable cross-talk bug" caused by an implicit type-casting error inside our background hardware synchronization loop.

To solve this, we ditched the anti-flicker slider entirely and replaced it with a clean 50Hz / 60Hz dropdown selection box. This wrote the frequency block directly to the registry, completely isolating it from the sharpness control.

(Fun side note: Right in the middle of diagnosing this bug, my delivery guy showed up at the door with a new magnetic GPU sag holder! Always nice to get a quick hardware upgrade mid-challenge.)

The Verdict: Did It Work?

Yes, absolutely!

Against all odds, the final app turned out incredibly robust. It doesn't just work on paper—it actually functions as a daily driver on my Windows system. Check out what this fully "Search AI-built" utility can do:

Hardware Rescanning: Instantly detects newly connected USB cameras with a single click.
Direct Hardware Sliders: Live adjustments for Brightness, Contrast, Sharpness, and White Balance etc.
No-Interference Dropdown: 50Hz/60Hz toggles to eliminate light flicker.
Saving Custom Presets: Generates custom .ccp (Camera Control Preset) files. I can crank my brightness to look like a ghost, click my "KB Preset" button, and instantly restore my perfect lighting.
Properties Sync: An "Advanced" button that opens the native Windows camera properties panel. Adjusting settings there dynamically updates our custom app's UI in real-time.

Try It Yourself (Open Source!)

Because this utility is still fresh and undergoing testing, I haven't packaged pre-built binaries yet. However, the code is fully open-source, and you can easily run it from source in a few seconds if you have the .NET SDK (v8 or v9) installed.

Clone the repo:

git clone https://github.com/kil0bit-kb/WebcamPresetManager.git

Navigate & Run:
```
cd WebcamPresetManager
dotnet run
```

Check out the code, try it on your setup, and star the project on GitHub:

👉 WebcamPresetManager on GitHub

Conclusion

This challenge proved that you don't need expensive subscription tools to build useful software. If you're willing to roll up your sleeves, do a bit of manual file creation, and copy-paste through some errors, the free AI tools available inside Google Search are more than capable of helping you build great utilities.

I’ve also made a video on this topic — you can watch it below.

https://youtu.be/jP5lUJukQrg

Check out my other posts, I post useful tutorials and tech tips, maybe you will find something useful 😉.