Google AI Studio

Google AI Studio is one of Google’s equivalents to ChatGPT, in that it is a chat interface on top of their proprietary Large Language Models.

It is slightly more targeted at developers, and offers a certain amount of free usage of the Gemini models with a Google account.

We’re not going to focus too strongly on AI Studio itself, instead just using it as an interface to the Gemini class of LLM models from Google.

Improving the Software Spec

Now that we’ve generated a software spec, we’d obviously go ahead and share that with our team and stakeholders to find out what’s missing, or if any other questions arise.

But we don’t have a team. So instead we can ask another AI to review the spec that we’ve created, and use that feedback to address and issues.

In the video we’ll be using Google AI Studio to iterate back and forth, but there’s no real reason why you can’t stick with the same LLM, just be sure to start a new session so that it doesn’t “remember” that it’s working on it. In ChatGPT, it’s sufficient to start a new chat.

This “duelling-AIs” technique is a really good way to refine a software spec, before you then review it yourself and send it out. Somewhat surprisingly, it does quite a good job of fixing its own mistakes.

Writing Code

As previously mentioned, writing code with a chat-only interface is quite a lot of effort. Although the underlying models can create the code, as we’ll see in the video, you’d then have to manage taking it back to your project, creating the right files and integrating it manually.

But that’s not to say it doesn’t have a place. To get code generated in AI Studio or ChatGPT, you just have to ask, and it’ll spew out endless lines, in whatever language you’d prefer.