I had issues with a LTE modem in an old Toughpad that wouldn’t work under Win11 but would under Win8. I gave it to GPT-5.5 and it built a new driver for Win10 that worked properly, using the datasheets for the modem from the manufacturer.
Vendors releasing proprietary drivers and software for their hardware should probably just opensource it. It’ll save the world precious compute. If they don’t, many customers will waste that tasking their LLMs to reverse engineer their crap.
I recently upgraded my PC and bought a case with a segmented display. Required a windows driver to work. I also bought a CPU liquid cooler with a display on it. Also required a windows app.
I threw about 30 bucks in tokens and now I have linux ports for both and with every added every feature I could come up with.
Now any weird or obscure piece of hardware I have that has been defunct for lack of software support has been getting the same treatment.
This is very interesting and I feel an example of a positive use of genAI. Woruld you mind describing your work flow to make this happen?
I used the github copilot tui (since that was what my subscription was through) and told it what the devices were and what I was trying to accomplish. From there it was a mixture of different things to figure out how the protocols worked, like running the apps in wine and having it log and capture data as I changed settings, or have it flip bits and I would report to it what changed. If it discovers a component in the device is something off-the-shelf, have it find reference material from the component manufacturer.
As you figure out how a particular protocol works you have it document it until you have everything you need to build your own CLI to interface with the device. At that point you have a pretty good idea of how the device works and it’s capabilities and you can build out whatever features you want around it.
I threw about 30 bucks in tokens
Dude, seriously check the subscription plans, unless you’re using your work tokens and they don’t mind.
My work pays about 100x as much per token as a $20/month subscription plan charges when you use all the subscription.
Oh I do pay for a subscription, just blew through my budget.
I have since moved to using mainly local models for everything. Running AI was why I was doing this PC build in the first place, just didn’t have it fully configured and I am wildly impatient when I get new stuff lol
I upped from the $20/month subscription to $200/month for a few months, it was pretty hard to run out of tokens on the $200/month plan - and have any life away from the keyboard at the same time.
I’m almost afraid to post about the various weird-cheap hardware I’m using now because the LLMs are setting up the cross compile environments and other PITA details that I didn’t have the time for in the past. I stuck with Raspberry Pi like a lot of people, and now you can’t buy a Pi Zero 2W for $15 except when the blue moon is on a Tuesday - don’t want that to happen to my newfound accessible toys.