thanks for remind me this. The behavior you mentioned sounds strange for me because when I asked Claude to enable the audio from the Waveshare display that I've bought for the FreeBSD based phone that I'm building,it offered me the DMA as a better method for producing a clearer and stable audio path. I simply accepted and now I have it. Maybe because I've got the habit to ask it to read how has been written the Linux code before to start writing the implementation for FreeBSD,so it is inspired by the code that you offer to it as a starting working base and it takes the working code base as a model to be inspired to make a new implementation. Acts this way and let me know.
I think I am not being clear. Just because frontier AI models can throw a bunch of low level code (regardless of how complex the project is) and the code runs without any errors, it absolutely does not mean your code is useful. It is actually the complete opposite. 9 times out of 10, devs would have to hand write your entire AI generated code from scratch because it is not fit for production or trusted public open source.
Do you realize that mission critical tech sectors and Fortune 500 companies all require and use memory level I/O workloads? Do you know that they usually use open source tools and projects which are only the most respected, trusted, and secure? What makes you think they or we are going to use your project which is entirely generated by AI? They have tons of money and spend it to hire smart people to get the job done right so that millions or billions will not be lost from their revenue.
Also, do not assume that frontier AI models are going to remember every single thing you feed them. Eventually you will fill up its maximum memory history capacity allowance. There are numerous papers now being released by computer scientists to find a solution to this. Not a single one of them has become a standard, and every AI tech company has their own way of doing things which always ends up being limited or very expensive.
I highly doubt Claude will store and remember every single project you feed it and will continuously parse through it to assist with every new prompt or project. The staggering amount of compute resources required for this would make Claude unprofitable (unless paid users pay tons of money), and its prediction performance (intelligence) would also degrade because its context window must always include massive past events.
Do you want ready-made "food" ? The fact that the code produced by Claude is not optimized is a valid reason for a programmer to optimize it with its help. I think humans should cooperate with AI,not that they let them do all the work OR that they don't want to be helped at all by AI. Every extreme position is not good.
Well, this is all I wanted to hear from those who solely rely on AI to do work. Clearly AI is not capable of getting anything done perfectly. This proves my point that we still need to review and babysit the code it gives to us. If none of the code it provided has been reviewed and optimized, I do not think devs would like to play with it or integrate it into real or paid applications.
If you are paying for AI generated code from your own pocket, you should make sure that other devs really appreciate and value the quality of your work, or else we would be better off doing everything 100% from scratch ourselves the right way and never touch your code. That is essentially a waste of your time and money.
I asked Claude to explain me how could I have replicated the functioning of VFIO on FreeBSD. That is, to write an implementation as close as possible to VFIO for Linux. Simple. It offered me ivshmem.
No, as a low latency professional firmware developer, nothing is just simple, if it was, professionals wouldn't be getting paid well over $100K a year to write code for high memory performance and efficiency. Speaking of today's memory RAM market crisis, its even more critical that the low-level code is at it's most peak highest optimal performance.
Using frontier AI since 2022, it never does anything professional. Things like this require mandatory professional work. The work you are doing involves critical memory level objectives, and it requires serious attention to detail and optimization. You must know what you are doing and why you are doing every step of the way.
When AI does most of the work, that is 100% trolling. We can assure you whatever code Claude or AI provided, 70% will have flaws in so many ways in the eyes of professionals. Talk is cheap. I want to see the code or the details and specifics (if it is open sourced).
Every hardware, low level, and firmware engineer's worst nightmare is when a project's entire code is written by AI LOL.
www.qemu.org
