The project that made me rediscover the joy of building Sireno Assistant: An AI assistant (of course) for forms

This article was published originally in Fika.

These last few weeks (I would say months), I thought I had lost the excitement and the desire to build things, which is what I’ve been doing for the last 30 years. But after hitting rock bottom, I realized that wasn’t actually the case; it was simply that what I was doing had no value and wasn’t motivating or exciting for me.

During these last few weeks, I’ve had time to reflect and share my thoughts on social media to try and start some debate, and I found myself copying the text I had written on a given page, taking it to whichever ChatGPT was at hand to ask it to correct my spelling and, in some cases, help me improve the claim. And then copying the result back to the page.

That’s why it occurred to me that it was a good opportunity to try a new development workflow I had pending, and using OpenCode I developed a Chrome browser extension. It simply allows you, by clicking a button added next to the field where you are writing, to interact with that field and an AI (OpenAI, Anthropic, Gemini, and soon Groq) to do exactly what I mentioned before (and more) without leaving the page you’re on.

Something as natural as this:

 

Of course, I got really excited, because I was enjoying myself like a kid again, building and imagining the product from the user’s point of view—from the perspective of the person who will use it—because empathy when developing and thinking about a product is very important. Just like diversity, which allows for different viewpoints and breaks many of the assumptions and biases we humans naturally have (and now I’m getting sidetracked).

As I was saying, in the end, I added some more features, such as the possibility of using SKILLS, even limited by domain. For non-technical people, SKILLS—that buzzword from the last week in the AI world—is essentially just text that tells the agent (AI) how to behave or what to do at a certain moment. For example, you could have a SKILL that contains some of your personal data, at least the basics you use to complete registration forms or similar, so when you ask the agent: “Fill out this form for me,” you don’t need to explain anything else, because it already has that information.

Obviously, this tool should be seen as a prototype; it still has many small bugs (elements that don’t position correctly, loss of some formatting, etc.), but it is a functional prototype. So much so that I’m using it daily, and I’ve decided to publish it in the Chrome Web Store: https://chromewebstore.google.com/detail/sireno-assistant/ppnecnomfnadhbcpacgkfldiapoejogo

If you find it interesting, I’d appreciate it if you tried it out (you just need an API key from OpenAI, Claude, or Gemini) and told me what you think (no filters), what bugs you’ve found, or what else you’d like it to do.

For me, this “project” has been something therapeutic and healing, so just having built it has already made it worth it.

To finish, as an anecdote, I already used the name “Sireno” in the past to name a CSS framework that used CSS Grid when it was added to Blink and WebKit-based browsers by the Galician company Igalia, which I called Sireno Grid.

“El Sireno” is a sculpture located in the Puerta del Sol in Vigo, and although I consider it aesthetically, let’s say, not very pleasant to look at, it is already part of my city’s identity.