Hey yall - so inside the folder is a working and ported version of the Open303 clone (put only the squeal2400 folder inside the LV2 folder - the tar.gz files is for you nerds to examine) that i managed to port by using Manus AI (because I know absolutely nothing about coding) - it works but doesnt sound like a 303 at all - this is more a proof of concept that using Manus AI it is possible to port - well at least this opensource code (linux 86) to the Aarch64 format - in between small fixes have been made like taking away the GUI and the sequencer of the Open303 and fixing sustain - first almost no sustain then full sustain then gated according to the midi note .- which seems to me that you can instruct Manus AI to make changes within the code and spit out something that is usefull. Right now im trying to improve the sound and the filter of the plugin and various new things like selfresonance a proper 18 db diode filter etc - this has been such a huge pain in the arse - one of the problems being that you run out of Manus computing at a certain point and has to start over which can lead to offshoots that doesnt work etc . i am at a point where i ran out again but have saved everything in a pkg that should make Manus AI be able to start from where it left off .. to be continued - have fun - peace Squeal2400 - Google Drive
on a side note ive tried switching over to Chat-gpt - im on Chat-gpt+ - maybe im too much of a noob to give it proper prompts but it is a complete shit show compared to Manus AI - for now at least - one of the main things i think that Manus has acces to a ubuntu sandbox and can acutually deliver proper files that are saved in the chat stream where as Chat-gpt will promise you stuff that will tank right before delivery or holds almost blank files - so if you got Manus Ai save some of the computing power and use Chat-gpt for research plugin algos LUTs etc instead 
Well you are lucky to have an account with Manus. Still waiting for my invite code.
Have you tried asking manus to do the project in smaller increments that it can iterate on? I watch some YouTuber do that with an app he was building. Logically it makes sense, until they free up more processing time. I can definitely vouch for what you’re saying about ChatGPT - I’ve been trying to have it help me develop some software and it was led me down a rabbit hole of frustration. Mainly b/c I can’t fact check if the code it spits out is even accurate, since I am not a coder.
i feel ya with chat gpt - maan also led down various rabbit holes with it but also with Manus - im not a coder either - i think with Manus it depends of its acces to the sandbox which failed from time to time giving me resets. But yeah doing things slowly and incremental must be the way forward not at least saving logs along the way. Can you maybe link to the youtuber? 
Sure thing:
Thank you!
More proof that using Manus AI is feasible - I succesfully extracted and implemented the “drive” part from another opensource synth called Blackbird by using Manus. I made Manus create a little webpage that sums up the process. The big takeaway from this session is that i now have a package that i can dump on Manus so it can start from the last development point and and can create a plugin compile environment. This should speed things for upcomming sessions! The googledrive folder now contains the latest version - see first post (yeah i know i need to version the plug better but im too lazy rtn) s2400_plugin_development_final.mdx - Manus Page Next im gonna look at portamento styles inspired by the Calf Mono Synth - do something about the decay parameter, filter resonance, add subosc etc