Oh no! Where's the JavaScript?
Your Web browser does not have JavaScript enabled or does not support JavaScript. Please enable JavaScript on your Web browser to properly view this Web site, or upgrade to a Web browser that does support JavaScript.

AROS AI automated Hardware Driver Research!

Last updated on 7 hours ago
AROS-ConfidenceAROS-ConfidenceJunior Member
Posted 8 hours ago
That`s my suggestion to integrate AI in AROS!

thaank you!

QBit

Jorge

https://opencode.ai/

https://github.co...rabin/AROS
Edited by AROS-Confidence on 18-11-2026 06:28, 8 hours ago
C
coffeecatSoftware Dev
Posted 7 hours ago
Breathe in, breathe out... I don't know why I'm answering, but here goes.

OpenCode is a tool that lets you run AI coding agents using different models. They can work with your code, compile it, run tests, interact with VMs, and cross-compile for platforms like AROS.

AI can be extremely useful for writing software, as demonstrated by many people, including yours truly. My own setup lets me write code extremely quickly and run tests across v1, v11, and Raspberry Pi in parallel.

But there's a difference between porting a game from the late 90s or early 2000s and developing hardware drivers. In the former case, I don't really have to worry about the hardware. With drivers, that's where the fun starts.

You can't simply tell AI: "Hey, here's a Linux driver, port it to AROS. Can you have it done by noon?" Well, you can, but good luck with that.

To develop a driver properly, you ideally need:

- Documentation of the hardware you're targeting.
- Knowledge of the OS driver architecture, APIs, interrupts, DMA, etc.
- Existing drivers to use as references, e.g. from Linux or BSD.
- A setup where you can actually run, test, debug, and repeat.

For example, I'm planning to experiment with drivers myself. One possible setup would be a dedicated machine running Linux and QEMU, with PCI passthrough for a network or graphics card. A second machine would run the AI agent, control the tests, collect logs, and observe the results. You could even add a webcam if you wanted AI to see what's happening on the screen.

Of course, passthrough won't work for everything, but the idea is to give AI a way to test its own work against real hardware instead of just guessing.

OR... you could approach this the way Haiku did. They built bsdcompat, a compatibility layer that allows them to reuse BSD network drivers without rewriting everything from scratch.

Something similar might make sense for AROS, but that would require some architectural and philosophical decisions about where we want to take the system.

So yes, AI can help enormously with driver development. I'm planning to use it myself. But giving OpenCode a link to the AROS repository isn't quite a driver development strategy.

But with that said - try it out and come back with the result. Port one driver and let's try it out!
T
terminillsMember
Posted 7 hours ago

coffeecat wrote:

@coffeecat - Breathe in, breathe out... I don't know why I'm answering, but here goes.

OpenCode is a tool that lets you run AI coding agents using different models. They can work with your code, compile it, run tests, interact with VMs, and cross-compile for platforms like AROS.

AI can be extremely useful for writing software, as demonstrated by many people, including yours truly. My own setup lets me write code extremely quickly and run tests across v1, v11, and Raspberry Pi in parallel.

But there's a difference between porting a game from the late 90s or early 2000s and developing hardware drivers. In the former case, I don't really have to worry about the hardware. With drivers, that's where the fun starts.

You can't simply tell AI: "Hey, here's a Linux driver, port it to AROS. Can you have it done by noon?" Well, you can, but good luck with that.

To develop a driver properly, you ideally need:

- Documentation of the hardware you're targeting.
- Knowledge of the OS driver architecture, APIs, interrupts, DMA, etc.
- Existing drivers to use as references, e.g. from Linux or BSD.
- A setup where you can actually run, test, debug, and repeat.

For example, I'm planning to experiment with drivers myself. One possible setup would be a dedicated machine running Linux and QEMU, with PCI passthrough for a network or graphics card. A second machine would run the AI agent, control the tests, collect logs, and observe the results. You could even add a webcam if you wanted AI to see what's happening on the screen.

Of course, passthrough won't work for everything, but the idea is to give AI a way to test its own work against real hardware instead of just guessing.

OR... you could approach this the way Haiku did. They built bsdcompat, a compatibility layer that allows them to reuse BSD network drivers without rewriting everything from scratch.

Something similar might make sense for AROS, but that would require some architectural and philosophical decisions about where we want to take the system.

So yes, AI can help enormously with driver development. I'm planning to use it myself. But giving OpenCode a link to the AROS repository isn't quite a driver development strategy.

But with that said - try it out and come back with the result. Port one driver and let's try it out!


For native a nanokvm, Null modem cable and a dedicated machine is quite handy.

My AI has it's own PC for native testing it has all of a benifits of using QEMU but adds the real hardware quirks.

That being said qbit should be ignored.
Edited by terminills on 18-11-2026 07:08, 7 hours ago
You can view all discussion threads in this forum.
You cannot start a new discussion thread in this forum.
You cannot reply in this discussion thread.
You cannot start on a poll in this forum.
You cannot upload attachments in this forum.
You can download attachments in this forum.
Moderator: Administrator
Users who participated in discussion: AROS-Confidence, terminills, coffeecat