Is there anything stopping this from being thrown some investment plus LLM tokens? I wonder what a KVM connected to the machine plus a few thousand in Astra/Fable tokens could do here. It seems funny to think of software feature defects in OSS software now if people are building full recombinations of closed source software daily now!
Add a tool to the harness you use that can ping you on Telegram/whatever when it gets stuck and need you to do stuff physically, tell it to use it in the workflow with you. Used this approach a bunch with Codex to work on and debug hardware stuff. Once every 5-15 minutes I receive my command, and as a good r̶o̶b̶o̶t̶ human I moved the resistor/patch cable in the way the agent instructed me, eventually it got things working in the way I wanted it to.
The problem with missing drivers is not "typing code fast enough", it is knowing that you need to poke 5 into [base-address + $0562] within 5 milliseconds of receiving an IRQ or something stupid like that which good docs would tell you why and how, but most/many vendors will not hand out docs on how to program their chips at that level. They will have docs on how to paint rectangles on an existing surface using the gfx card, but not on how to actually start it at poweron or something along those lines. If you can't make it start painting a screen and waking up the monitor at boot, it is of very little use that you can paint rectangles in the dark.
So even if you can divine that a binary windows driver sometimes writes a 5 into that register at certain times, you will have a far harder time figuring out how and when it is needed to actually do that, and the source you get will be super hard for anyone else to understand if its all magic numbers getting written into "random" locations, compared to say, "wificard.driver.radio_enable = ACTIVATE_RADIO;" even if it compiles into writing 5 to offset $0562. Such reverse-engineered sources are super hard to keep working in the long run, when someone rewrites how and when IRQs are delivered to devices or whatever major is happening in the kernels so this leads to drivers no longer working even if they did have a short period of use.
I built this concept a few weeks ago even - not nearly as good though, I used wiki data plus LLMs for most of the data sourcing and reverse geo coding but the dataset here seems a lot cleaner.
Thank you! I think there are many existing competitors already when it comes to maps for finding movie scenes, so none of us are unique I guess.
I try to do my own thing and hope for the best. I rely both on my own principles and standards + Feedback from the community so that it gets quality tested many many many times.
> Meanwhile, donations and pledges kept slowly decreasing, and have done so since the project launched. Not enough to spell immediate doom for my dream of working on Asahi full time in the short term, but enough to make me wonder if any of this was really appreciated. The all-time peak monthly donation volume was the very first month or two. It seemed the more things we accomplished, the less support we had.
I think the issue is funding is so low the developers don’t even have access to newer Apple Siicon, let alone funding for token costs, etc.
> It is the opinion of the Board that Large Language Models (LLMs), herein referred to as Slop Generators, are unsuitable for use as software engineering tools, particularly in the Free and Open Source Software movement.
Using LLMs in a project like this is also somewhat illegal, because LLMs may have inside information about apple devices and it's not allowed to use it.
That's nonsense, then it would also be somewhat illegal to use it inside Apple, being trained on GPL licensed code (and it's not even a possibility here but a known fact)
Can you please edit out swipes like "That's nonsense" from your posts here? This is in the site guidelines: https://news.ycombinator.com/newsguidelines.html. Your comment would be just fine without that bit.
(I realize these little things may seem insignificant, but (1) it's customary for people to underestimate the provocations in their own posts and overestimate those in others, and (2) they add up like mercury in the bloodstream.)
That's not going to save them, the model could leak GPL licensed code on output as well, and it's an order of magnitude more likely than leaking internal documents.
These large tech companies can't have their cake and eat it too.
It's not new Anthropic/OpenAI/whatever trained on copyrighted code and leaks (anna's archive). Companies just don't give a shit because you can't prove it.
In Asahi Linux specifically, there was not long ago a PR for M3 or similar for removing a big roadblock that existed (and I think still exists). Turns out this whole PR was implemented by an LLM (using copyrighted materials, probably most of it is Apple's) and it was immediately closed. Yes, it's a problem and illegal.
If you can't prove it, you can't prove it both ways.
Either everybody gets to use it or nobody does.
Especially that the code getting trained on is for the vast majority copyrighted open source software so if anybody has something to say, it's the opensource community rather than Apple
But weren't they all made outside of Hollywood? Emeryville, around Europe and the UK, and then the UK for Toy Story 5, Odyssey, and Spiderman respectively?
Which is cool, but also likely an outlier due to Christopher Nolan's usage of practical effects vs CGI. His clout probably helped with the funding of that tbh.
reply