Skip to main content

AI usage

fuegocable aims to be a reliable source of knowledge and provide a solid, stable hardware and software solution for modifying audio interfaces. We’re not there yet, but LLMs can help us achieve this goal. First and foremost, by making technical information more accessible, finding datasheets, and helping us understand complex concepts. And, of course, by generating code. Above all, however, we will always prioritize human judgment and good taste, whether in documentation, images, or code structure.

At the moment (2026), the use of generative AI in the project is quite limited. The documentation was written by a human, as was practically all of the firmware code. This does not necessarily mean it will always be that way. Translations to english and code-review were done with help of LLMs. AI can be helpful and is a great tool, but for this project, we will not use it without supervision or in an uncontrolled manner, for several reasons, including the following:

  • Modifying a digital audio bus is a critical process when it comes to stability. As such, it must be approached responsibly. We need to understand how every part of the mod works to maintain control over it and ensure the highest possible audio quality and reliable operation of the modified interface.

  • The present is noisy, and the future is uncertain. AI seems to be here to stay, but the industry is changing so rapidly that it is worth taking a step back and approaching things calmly. Ultimately, we do this because we want to, and given how these devices are typically used, avoiding unnecessary haste is the most reasonable approach. There is no need to rush to deliver a new feature or merge a PR. Let's take our time.

  • There is a special kind of joy in doing something simply for the pleasure of doing it. This might mean hand-coding a new feature, making design decisions, or reverse engineering a bus. If there is a process we can enjoy, such as implementing a new feature or reverse engineering a data bus, why delegate it to an automated process? If you enjoy solving Sudoku puzzles, you still solve them yourself, even though thousands of programs can do it in an instant.

written by a software developer in crisis :)