Thread 106411842

Can an LLM convert binary compiled code into C?

>>106411842 (OP)
I don't know have you tried it?

>>106411842 (OP)
If it was trained to do that, then perhaps.

>>106411842 (OP)
in theory- yes
in practice- no
not enough context size
not enough data

>>106411842 (OP)
Why on earth are you creating a thread for this instead of trying it

>>106411842 (OP)
>>106411858
>>106411853
>>106411852

There were already some examples where using Ghidra MCP server LLMs have been able to completely reverse engineer and modify compiled code. It does take a bigger reasoning model, a RAG and a little work but it is possible

>>106411842 (OP)
ai sloppa

>>106412225
>and a little work
how much, exactly? hmm?

>>106411842 (OP)
>Can an LLM convert binary compiled code into C?
no

>>106412225
btw
>ghidra mcp
>an api for ghidra
*horseshit, not llm
ftfy

its vanilla fucking ghidra with a wrapper around it allowing it to be launched by the llm
its not the llm that does the fucking decompiling, ghidra does

>>106412315
>95% of enterprise AI fails
that's because these startups are nothing more than AI wrappers that contribute little to no actual discernible value compared to their competitors

What is the real difference between v0, replit, Loveable, Cursor? All of them don't have their LLM models but wrap other LLM providers alongside their infrastructure and templates. The reason most fail is because it is a hard field to get into and stay relevant, high competition drives high failure.

The amount of energy someone needs to spend using Ghidra MCP and Claude is relatively much smaller than doing it by hand. If they know what they're doing then it would be trivial to incorporate in their workflow without any issue. Add also STT and TTS and you got a programming buddy that can help you quickly on the fly.

If you want to stay a luddite please do, just don't drag others with you.

>>106412351
>>an api for ghidra
It's specifically made for LLMs to use it. MCP servers are APIs for LLMs. You query LLM, it queries Ghidra and does the job for you which speeds up the process tremendously.

Did you even see the demo mp4 on the github?

>>106412385
yeah but its still in no way the llm that does the decompilation
its like saying a wall was built by an architect
>the demo
idc abt the demo i know exactly what to expect and i assume its working

i also absolutely fucking despise ai cultists like you and your deliberate dialectical imprecision which you user to shill chatbots
i use chatbots too. i know exactly what theyre capable of, and what they cant do

i dont need a marketingroid retard from reddit to tell me what to think

>>106412385
>i use chatbots too. i know exactly what theyre capable of, and what they cant do
in fact, i use chatbots in the same way as used by the 5% who benefit from ai, as outlined in the mit paper
not as a specialist tool but as a generalist one, as basically an interface to the internet

>>106412412
>dumb retard starts talking about "muh dialectic" when he was first to start using dialectic

I ask you in return. Why even use Ghidra then? Why not use dbg?
Why even have an UI when decompiling? Why not code your own tool to do everything in memory?

If you don't want to use an architect, a crane, a wheelbarrow or even a shallow to build a wall and want to do it by hand then go ahead.
Again, I have no issue with you doing everything "the old fashioned way", I just want you to know that you will in few years time either come on our side or be left in the dirt due to your incompetence and arrogance.

At the end of the day your dismissal comes from fear alone. Fear that your incompetence and arrogance has put you in corner of irrelevance. And instead of your love for technology prevailing you get angry at anyone saying otherwise.

>>106412412
>>106412490
PS
I am an AI/LLM developer and I work about 2 hours a day because LLMs speed me up so much that other 6 hours a day I spend either shitposting here or doing my own stuff

>>106412490
>I ask you in return. Why even use Ghidra then? Why not use dbg?
i used both. have you?
i doubt it bc you wouldnt ask me a question with such an obvious answer
its different tools serving a different purpose

>At the end of the day your dismissal comes from fear alone.
nice sad homonem but im a solodev.
i would 100% benefit from being able to offset parts of my work to chatbots
and youre a fucking retard, anon
you cannon even conceive people can be self employed
its safe to assume your opinion on other things will be equally worthless

I actually had this thought, but with something like ghildra or ada pro. It would be cool to have an extension that took the decompiled code and automatically renamed the generated variables and function signatures to more accurate reflect what they are doing. It would save so much time. It would be a cool use of a LLM

>>106412517
>I am an AI/LLM developer and I work about 2 hours a day
yeah, no, keep your mythomania in reddit

>>106412517
yeah im the same at this point, i try to get a vibe coder general going but it doesnt stick

>>106412549
maybe because its fucking worthless and everyone just hides your threads?

>>106412535
>i would 100% benefit from being able to offset parts of my work to chatbots
But anon, you would lose your bragging rights about being a 10x self-employed engineering genius anon, think about it
You can't use LLMs to help you in any way, that would imply that you're dumb, and you can't have that

>>106412559
I bet you never used LLM for more than 5 minutes 3 years ago and take truth from idiot normies

>>106412545
>I can't believe it, they must be lying
stay in your bubble luddite

>>106412562
>think about it
yeah no im not in the mood for dialectics
i could have been more lenient if this wasnt the 123456789# ai shill thread

if only your piece of shit chatbots could do half of what you advertise i would have been elated
but no
theyre chatbots and shitty ones at that
ai codegen has been solved years ago btw
but thats for aerospace

goyim get to play with the dyslexic stuff

>>106412599
>he doesn't want to think about it
you sound like indie game dev not gonna lie

>>106412599
forgot picrel
i lack words to describe the levels of ignorance you operate at

>>106412605
i told you its a solved problem, retard
what the fuck is there to think about?
>yeah but if x and y and z gave a blowjob
thats creative writing
yes, you "write" a program too, but its not the same type of creativity

>>106412626
You do really have a high opinion of yourself don't you, I wonder if its deserved

>>106412640
>more talking about me
im very flattered by the attention but i know myself quite well
tell me something new
tell me something related to the tech were discussing, retard

>>106412660
I don't think you do

>>106412740
>more talking about myself
is it because you have fuck all to say about the technical aspect of things?
this question is purely rhetorical

at this pace im learning more about you than youre telling me about myself

>>106412794
ok sure

>>106412815
well, youre fishing for bites
but in the process youre exposing yourself
because i can deduce what are your assumptions
which in turn allow me to deduce a lot of things, like your system of values and lived experience

which in turn allow me to deduce things like that you probably arent from here because you assume im a gregarious person
which tells me you dont post here often because you would know were pretty individualistic in this neck of the woods
and that you come from a place where gregarity is the norm
because ours is not the only neck of the woods where people pride themselves on being individuals

>>106411842 (OP)
can you?

>>106412893
Wow, this scene. Gonna have to measure how it ages pretty often aren't we?

>>106411842 (OP)
toxonigs really are one of the lowest IQ retards online