In my last year of high school (for those familiar with the Indian education system, that's Class 12), for my final project I made an image compressor/decompressor in Turbo C++ that used a discrete Fourier transform (DCT, actually) to work on 8x8 px blocks of images. I based it off some stuff I found about how JPEG compression works. It worked even better than I'd hoped even though it was slow as molasses (I programmed a naive Fourier transform instead of using a FFT variant). I remember jumping with joy all around my room at 3 o'clock in the morning like an excited gas molecule in a box. Fun times.

My last post was a year ago. What brought me back here is the ability of AI to agree and apologize to anything and everything, while producing the worst hopeful code.

4 days I wasted, trying to make an android audio visualizer, but AI... sigh.
It gave me the wrong structure of FFT bytes emitted. I corrected it
It gave me the wrong logarithm calc, I corrected it
It gave me the wrong sampling rate, I corrected it.
It gave me the wrong texture order, I corrected it.
It gave me the wrong glsl sample2d, I corrected it.
It gave me the wrong textureID generation, I corrected it.
It gave me a render which was about 10 fps, I found out that instead of using native onDraw, I had a fcking delta time in my shader. I almost corrected it, I gave up

Lets go to code generators with Annotations.
Like always, starts very positive, until I start to correct it.
It gave me the wrong file locations, I corrected it.
It gave me the wrong order of find copy modify and write to .build, I didnt correct it.
It gave me regexes to find annotations. Im like So whats the use of an "ANNOTATION PROCESSOR"
It apologizes and used a fucking regex in the processor,..... I didnt correct it, in the end, I was left with a separate module, targetting iOS Android and JVM, with an annotation processor implemented in jvmMain, which tries to modify commonMain src by finding annotations with regexes, which wont run on app build or app sync project, but only on java -jre command pointing to that fucking .java class in that module, which takes at least 2 mins to run, and Finally generate 0 files.

I needed to rant, I understand LLMs are just models of words built and stolen from the most intelligent and dumbest people out there. But Im an idiot for getting my hopes high. I cant build anything new and unheard of. I used to do that. I once made a textView + image print util for a bluetooth printer just to say FU to libraries and heavy sdks. like literally rasterizing shit to bluetooth packets. I needed to let off some steam. I havent been here in a year so I dont know what reactions I can get from this rant. I bet someone will just say yeah we tired of 'Fuck AI' rants. but shit, it hurts. When I gave up on that visualizer, I downloaded an app, I think its called project M, like in reference to MilkDrop.. like the Winamp Milkdrop. I opened it, played something on spotify, and let my eyes go blind

I have to implement DSP using Convolution, FFT. And for the past 2 days I have like a million tabs open in Chrome and half of the titles end with ".... for dummies" :D

At least I'm learning something new and when I understand what the math behind those terms does I feel awesome :D

Multiple projects on fire. WHY, GOD, WHY?

I am looking for a library in JS for signal analysis purpose , fft , graphs etc . Any suggestions