Vibe Coding Log - Sharing Journey - Git - Implementing more MIT delays

[Seafire Mk2]

Seems strange to do what he is doing. Everyone has an idea of their 'perfect' synth, but to go for that from the off, seems bizzare. If you know nothing about coding, you would think it would be a good idea to try something simple, instead of diving into the deep end. At the end of the say ( love a cliche) if you cannot code, you will always be stuck at some point.

Personally I hate these 'do everything' synths, they sound bland as hell. Much prefer something that is niche and has a great sound, but also requires the user to make an effort to appreciate its quality.

AI coding is a joke, and I'm pretty sure there are no devs are worried about it in the least, except Cherry Audio

[Seafire Mk2]

The important part is the code.

The bit he can't do

[Touch The Universe]

So much work.

(sorry, that one was too hard to resist)

Isn't it interesting, truly interesting, how the shape of work differs, but it's work nonetheless. By all means, it is work indeed to put together these ideas, many parts need to come together. It doesn't build itself, dream itself, though it does all the building in code, yet there is a building of ideas.

Ideas are work, prompting is work, piecing together ideas is work.

Just not the tedious kind, yet still tedious. However, I can only imagine the work involved coding these things from scratch. This is a few hundred hours of work. I imagine from scratch this would be over 10,000. Boggles the mind Not sure I have the patience for it, though it truly does seem to be rewarding work. Even having ideas come to life before my eyes is not always the motivation to grind through. i can only imagine the patience invovled coding from scratch. in fact, I don't believe I can, as there must be some intrinsic joy in that which I have yet to discover. Perhaps if I do learn, then I'll gain a double joy, and greater sophistication in prompts as well.

The future is scary. Are you afraid of the future? I am. Because the implications are of these things are biblical, in the worst kind of way. Today it's causing a crazy sound designer to build a dream synth, tomorrow, who knows what temptations it might offer. Throw great temptations inside a system of great corruption and great power, is but madness.

Imagine if you spent the 10,000 hours of coding and exchanged it into 10,000 hours of refining ideas in piecing together already existing maths and algorithms. Ideas and maths themselves become the macros of lines of code and the on the order of business is arranging them, like going from the microscopic to the macroscopic and dwelling there only. That's my world and it's quite amazing and quite a bit of work. I don't care what anybody says. Work is work, pain is pain, Joy is Joy, tedious is tedious. Failure is failure, whether you are dealing with ideas or codes. To me, that's truly interesting.

[Urs]

I'm not particularly scared of the future. It's interesting to see where/how people go with all that AI stuff, like you do, but I don't see anyone being able to build a proper business from it that could compete with my own. For instance, you reference a lot of open source stuff, probably GPLed, so you'd probably have to open source it. I do fear that other people use AI to circumvent the licenses of open source stuff, which IMHO gives them an unfair advantage (I sometimes feel that AI is a "license washing" business). These are not just risks for me but also for the people who do this. They might not even know that they are infringing on copyrights because the AI might have conveniently forgotten to tell them.

Software development itself is quite addictive. There's a lot of flow to be had, which IMHO I don't quite get when using AI.

[Almondo]

It's hard to imagine AI ever really competing in this area. I said it earlier in this thread, humans are forward thinking whereas AI is literally stuck in the past and always will be. You can throw all the ideas you like at it but it simply doesn't have the ability or data to create something unique. And the chances of Urs giving up his codebase for model training is about 0%.

Having said that, and as a software developer of 30 years, it's an interesting experiment.

[Seafire Mk2]

AI can only output based on what has already happened. Even then, its an approximation.

Real devs have nothing to worry about.

[Tiles]

I have to admit that I'm quite happy with how AI speeds up my work. My current projects are completed in days instead of weeks or months. The research alone can be done by simply asking an AI, rather than spending hours or even days crawling through API documentation or Stack Overflow, for example.

Real developers have always used whatever tools speed things up. But programmers do have something to worry about now. AI is a translation layer, just like traditional programming languages are translation layers between humans and the hardware.

[Touch The Universe]

After taking a break and meditating a bit on this while running, I've decided to revert back. I celebrated too soon. I was jumping for joy seeing a 70% reduction in the cpu go to like 12% but the win only came from bypassing the voice loop entirely, so all envelopes, fx, matrix, and ability to layer other synths were compromised. I essentially turned the entire synth into just OSC 1 page. I isolated this page alone and it is it's own synth for now so I preserved this low cpu version, in this form to test, but it broke too many other others I don't want to untangle. Same with the Vital OSC page, which is named that because its a facsimile of Vital OSC, not the imported code. GTP won't let me import that so casually into the synth, warning, after warning, etc. Good, yet, a bit annoying when i just want to test. I was finally able to test the real Surge FM code for ONCE, actual 3rd party code, but it wasn't easy, then removed it.

I also experimented with adding multcore support to a lot of success. I will try this route as well, eventually, after I get PERHAPS the explicit audio routing paths straight, though I'm not convinced they are all that broken, as I can layer the synth engines, hear the FX, so clearly there is explicit paths already in place or it wouldn't work. I will study more the paths, fix obvious hacks, like bypass filter on this page willy nilly, with no plan, and learn a bit more on this. The routing for gui, paramter id's etc and a few other things are also concerning me.

Regardless, is it normal to bypass the voiceloop alone and get so much cpu efficiency back? Obviously something MIGHT be broken, but I'm not knowledgeable enough to address it, yet.

I will focus on problem cpu areas, the FM, MODAL, and Reverb engines and do all I can to lower cpu, then make paths more explicit if They aren't already, but I won't request blind, I'll get proper paths before hand on what i have now, then see what is needed to fix. Some areas might be hacky, simply fix these areas.

Add Mutlicore support for "safe" things, like unison voices, while leaving more complex things perhaps on the default main core.

The goal is to be able to layer all synth layers and a few fx or there is no point to have a monster synth, or atleast half the point is missing, as i can still use it to have a wider breadth of sounds, it would be cool to be able to layer them, that's what Im after - or what's the point. That where the sound design really interests me.

I am treating this project as an extension of sound design, not vsti development, that is incidental, oddly, and I can have a synth to focus my efforts here, while having the door open of ZERO LIMITATIONS. If i don't like the comb, i don't have to make a wishlist request to a dev, I can just spend a few hours and add it in there.

Nonetheless, many ups and downs the last few days. Tiles was right, then I was, then oops, no, really dumb solution, haha.

Humbling little adventures of refractoring continues, but at this point, Im not convinced that is the PROBLEM area. I need to dig around more. AI convinced me it might be, but even if its poorly designed, I can see if I can save cpu elsewhere.

Eventually, I'll page by page improve the routing to make sure it's efficient, but I need to SPELL these things out in advance on the PLAN explicetely to the GTP so he can make the routing explicit, haha, or it's just madness.

I'm definetely feeling I need more insight, if I could just look at the code a bit and see the routing, that might help, or right now, feeling kind of blind. Hopefully I can get more insight on the ACTUAL state of the routing and not just guess it's good or bad.

[Touch The Universe]

I'm not particularly scared of the future. It's interesting to see where/how people go with all that AI stuff, like you do, but I don't see anyone being able to build a proper business from it that could compete with my own. For instance, you reference a lot of open source stuff, probably GPLed, so you'd probably have to open source it. I do fear that other people use AI to circumvent the licenses of open source stuff, which IMHO gives them an unfair advantage (I sometimes feel that AI is a "license washing" business). These are not just risks for me but also for the people who do this. They might not even know that they are infringing on copyrights because the AI might have conveniently forgotten to tell them.

Software development itself is quite addictive. There's a lot of flow to be had, which IMHO I don't quite get when using AI.

For your business, of any dev, I also would not be scared, but I was mostly speaking eschatologically You being interested and not rage baited speak alot imo. As I said previously, I'm much more interested in vibe coding for sound design potential that I can make and sell banks, and for the sheer potential of this is hard not to be tempted. I'd be happy if I could make enough with vibe coding, it is hard to find myself not hoping for this, but nonetheless, as you say, I'm under no delusions on what is required for a successful business, namely marketing, promoting, as the least - building decades of history and relationships in the audio community, I'll likely never replicate, nor your success, but for me that isn't necessarily what I would consider success so to speak.

Speaking personally, I have my sight set on different areas, like real estate, as that seems more rewarding, yet nonetheless, there are dangers of prosperity, spiritually, which itself might not be a "success" if it pulls you deeper into this world, in ceratiain areas of distractions, and where deeper more life giving life fullfilling spiritual things take a back seat - as the proverb says, what does it benefit to gain the world but lose one's soul.

There is no 3rd part GPL2 code in my synth. I have OSC Vital to reflect the facsimile it's designed around. Whether this is getting around GPL licensing with GTP, is an interesting question. Math is math, what seems to be the licensing is the implementation of that and gui work, as far as I can tell. There is doubtless more involved than this. I'm just pleasantly surprised so many people just release there code for free. Vital is an outstanding resource. If everyone did this, synths would be better, and so would sound design, and my premise or theory would be that would NOT hurt any sales, so long as there is neutral playing fields of marketing, though there seldom is. This is because no one cares about amazing ideas more than the person who came up with it. You can have amazing ideas on your plate and place them on every ones plate. What causes action and perserverance are different than latching onto potentialities, ie open source. Otherwise how come we dont SEE any vital beefed up facimiles 1 to 1. How come no one has compiled and gathered all this code in a monster synth to sell, even at GPL, and open source it.

I'd like to see data on open sourcing increases competion and less sales. I think the issue is more deeper than that. But what do i know, just a sound design hack

In my experience, GTP ACTIVELY PREVENTS implementing GPL stuff in a project deemed commercial. It will pushback, ignore, make a facsimile without telling you, otherwise go at great lengths NOT to implement it. Perhaps because of the very issues you bring up. 20 times I tried to get them to implement REAL code to test, but got sidetracked implementing other ideas. I noticed the pattern, but just ignored it until it became pressing and I had to lay my foot down, insist, four or five back and forths and it was done, only after explicetly stating it's temporary build and we will revert, etc.

When I think of flow, i think of sound design, making music, or running, or maybe even listening to music. If coding can get into a flow, that is a joy indeed, added unto the joy of testing ideas.

I was working in zebra 3 and messing with the pattern vertical slider for the grain patterns in the texture module. That is a good example of an idea creatively implemented that is unique that I love and try to add similar things, where a knob is doing multivalent things.

I'd do a test if I was you, compile as many ideas as you can think, the creative sort like that slider, and have gtp implement them all at once. Work from this 20000 feet overview, dealing only with ideas, at the macro level, get into the flow of them, and have AI do all the implementing. Do a test run specializing in ideas, and not code, maybe then you will be afraid, not for your business, but spiritually. I'd imagine things to go more smoothly for coders because they can prompt explicetely.

I'll say this, I've sound designed in dozens of synths and spent 10's of thousands of hours sound designing. I am a decent judge of character of quality in a synth and this crazy monster synth that just manifested is nothing to sneeze at, at all. It's powerful and sound good. I don't know what to think anymore. It's 70% there. The feedback in the Legend is far more subtle, sophisticated, and a few behaviors in filters are not there, but as far as filters, distortions, fx, grain implementation, modal synthesis, fm, it's rock solid. The feedback I eventually had it spit out in different versions wasn't as polished yet somehow has a greater sonic potential for sound design.

Nonetheless, I appreciate your interest. Thanks for stopping by

[TechHaus]

Are you dyslexic (not a diss) because you keep saying "GTP", every time?

(i'm autistic and my brain stops me from reading every time I hit you spelling it that way)

[Touch The Universe]

Are you dyslexic (not a diss) because you keep saying "GTP", every time?

(i'm autistic and my brain stops me from reading every time I hit you spelling it that way)

I'm confused. Chat GTP is the agent. GPL is the licensing.

[TechHaus]

Are you dyslexic (not a diss) because you keep saying "GTP", every time?

(i'm autistic and my brain stops me from reading every time I hit you spelling it that way)

I'm confused. Chat GTP is the agent. GPL is the licensing.

It is Chat GPT.

[Touch The Universe]

It's hard to imagine AI ever really competing in this area. I said it earlier in this thread, humans are forward thinking whereas AI is literally stuck in the past and always will be. You can throw all the ideas you like at it but it simply doesn't have the ability or data to create something unique. And the chances of Urs giving up his codebase for model training is about 0%.

Having said that, and as a software developer of 30 years, it's an interesting experiment.

like new physics?

Try giving it crazy ideas like make a knob with unique Feedback implementation that has resonance peaks part of the knob path, like first 40% then have fm from 50-70% that affects multiple parameters at once to sound good and work with feedback, have distortion in the, and link it to speak with a ZDF filter so there is cross communication. I've never heard this in any synth. Is this not unique? It can piece together and do anything IN MATH. Those combinations can be unique. See for yourself. Just say build this so i can test in html, no need to build and compile.

[Touch The Universe]

Are you dyslexic (not a diss) because you keep saying "GTP", every time?

(i'm autistic and my brain stops me from reading every time I hit you spelling it that way)

I'm confused. Chat GTP is the agent. GPL is the licensing.

It is Chat GPT.

I'm dsylexic then

In my brain, it's GPT, I think the GPL mentioning crossed it in my brain somehow, even though the letters are lined up.

[TechHaus]

Are you dyslexic (not a diss) because you keep saying "GTP", every time?

(i'm autistic and my brain stops me from reading every time I hit you spelling it that way)

I'm confused. Chat GTP is the agent. GPL is the licensing.

It is Chat GPT.

I'm dsylexic then

In my brain, it's GPT, I think the GPL mentioning crossed it in my brain somehow, even though the letters are lined up.

LOL

Ayyy, your brain is doing a lot of work the past couple of weeks. Lots to juggle.