I like to dunk on LLM generated crap, but even I must admit that some people who already had skill in decomp or who otherwise care about the maintainability of the end result and have a well researched binaries or debug symbols and are attempting binary matching can get very good results that are close to what a human attempting binary matching can achieve in a fraction of the time.
Vino is a driver that lets the Linux kernel talk to DisplayLink docks directly. It is meant to replace DisplayLink’s closed-source software and the open source EVDI kernel module that sits beside it.
Mike wrote it with help from Claude Code, and the first version reached the kernel mailing list in June, unfinished.
This, among many other projects is why I think it’s just smoke and mirrors, and overhype.
The AI does its AI thing in all contexts : smears the average of various things together and if it sticks, it sticks. It “assists”. Fucking great.
The AI is not a fucking genie, which is what this way of thinking or speaking implies it is. If you want to, you can contort this crap into giving workable results.
If it was “good in translating crap from one language to another” as in software langauges, it could take source code in Actionscript and turn it into Javascript. They are both derivatives of ECMAscript. The project above wouldn’t be “unfinished” the guy would have finished in over a week (giving time for testing, etc)
But it’s not. You need to coax it into manageable chunks, test and assess, etc etc.
By doing this, you are giving up fine grain understanding and control over the system architecture. Great, you achieved The ThingTM and it’s an unmaintainable mess.
The magic happens in between analysis and performance. The actual strength in AIs ARE the hallucinations/emergent behaviour, because that’s the part it specifically brings to the table.
Learning Metaprogramming is hard and takes time. Paying god knows how much money to Anthropic is cheap for now.
I really don’t think so. Various ports and such have taken fucking forever. I tried (admittedly 2 years ago) and got shit results.
Two years is a long time in the AI field. Major developments happen in weeks.
For example, the Oracle of Time remake was released for the Switch 2 on Nov 5th. Someone had a playable PC port the next day. https://www.xda-developers.com/zelda-ocarina-of-time-reportedly-already-runs-on-pc-emulator-just-hours-afterswitch-2-release/
please reply here : https://lemmy.world/post/54283786/26814079
I like to dunk on LLM generated crap, but even I must admit that some people who already had skill in decomp or who otherwise care about the maintainability of the end result and have a well researched binaries or debug symbols and are attempting binary matching can get very good results that are close to what a human attempting binary matching can achieve in a fraction of the time.
context : https://lemmy.world/post/54280972?scrollToComments=true
This, among many other projects is why I think it’s just smoke and mirrors, and overhype.
The AI does its AI thing in all contexts : smears the average of various things together and if it sticks, it sticks. It “assists”. Fucking great.
The AI is not a fucking genie, which is what this way of thinking or speaking implies it is. If you want to, you can contort this crap into giving workable results.
If it was “good in translating crap from one language to another” as in software langauges, it could take source code in Actionscript and turn it into Javascript. They are both derivatives of ECMAscript. The project above wouldn’t be “unfinished” the guy would have finished in over a week (giving time for testing, etc)
But it’s not. You need to coax it into manageable chunks, test and assess, etc etc.
By doing this, you are giving up fine grain understanding and control over the system architecture. Great, you achieved The ThingTM and it’s an unmaintainable mess.
The magic happens in between analysis and performance. The actual strength in AIs ARE the hallucinations/emergent behaviour, because that’s the part it specifically brings to the table.
Learning Metaprogramming is hard and takes time. Paying god knows how much money to Anthropic is cheap for now.