Remix.run Logo
▲ timbert 3 hours ago

I think we need a new model of software delivery.

Here's my most-recent real-world problem: I wanted to plug in my midi controller keyboard and just play a piano, so I can practice my choral music. Very basic.

This turned into many hours over several days trying multiple DAWs, trying to get tehm to install, get them to run at all, get them to recognize my controller, get them to play a VST plugin...changing system settings, completely jacking up my sound, restoring it all again.

It was a nightmare.

So instead I fired up Claude, worked with it for about 3 hours, and now I have an *8-channel* VST player that supports all my hardware, and includes a snappy UI, with outstanding performance and sound. It's amazing!

I literally spent less time creating a custom solution than I did trying to get the off-the-shelf stuff to work. I could make my player OSS too, but the effort to do that is huge, for all the reasons this forum would understand.

Bespoke software avoids all of that and allows me to do what I want, with the environment I have. I don't have all the features of an FL Studio, but I have what I need, and that works for me, beautifully, at low cost and effort. What's not to like?

So maybe we should share OSS "templates" instead of full products, projects that are the barebones of a solution that you can clone, and have your LLM modify to suit your needs, on your platform of choice.

For example, perhaps my barebones VST player is mostly good enough for you, but you also want to play sound fonts? Cool, clone my project, instruct your LLM to add sound font support, and add support for your controller or your output device. Or convert it from Gnome to KDE.

The time and cost of doing this is so little now, why not? I'm getting more out of my computer than ever, and I'm happy about that :)

▲nayroclade 3 hours ago | parent | next [-]

Yes, right now this is absolutely the best way to build bespoke software, but it is also very wasteful. Think of all the time and the billions of tokens being burned to repeatedly reimplement apps and features. LLMS code fast, but not yet instantaneously or perfectly for complex features.

What we need is more than just OSS templates. We need fully LLM-managed "software libraries". These would be like package managers on steroids. They would be a huge hierarchy of LLM-coded modules. When you ask for an app, your LLM assembles it from the modules that already exist. If you ask for a novel feature, it codes it up and places it into the library.

So, for your VST player example. You would ask for a barebones version with just the features you need, and your LLM would assemble it for you from preexisting library modules. Then you ask for sound font support. If there is an existing sound font module, it would download it and add it into your app. If there wasn't, it would create one and put it into the library, so the next time someone asks, they can quickly get a VST player with sound font support.

▲jhloa2 3 hours ago | parent | next [-]

I have a similar experience where I’ve been learning the keyboard but didn’t like the available free options and the good paid apps had really expensive subscription plans so I had Claude build me something that runs natively on my iPad, with midi and microphone support and generates custom learning plans for midi pieces that I import.

I’m an engineer but not a software engineer and it’s been pretty interesting trying out vibe coding and seeing the barriers of entry being blown apart.

I’ve been progressing at learning the piano quite a lot faster than I expected with this. I’ve been debating open sourcing it but I’m not sure I want to take on the work of managing it

▲compacct27 36 minutes ago | parent [-]

Oh wow, just did the exact same thing on my end. If I’m struggling with some fundamentals I’ll just have Claude scour the internet for the best instruction and throw it into a lesson

▲red_admiral 2 hours ago | parent | prev [-]

This would work if all LLM contributions were not only open-source, but also public domain or something. Otherwise you'll get a mess of incompatible licences.

▲montenegrohugo 3 hours ago | parent | prev | next [-]

Fully aligned, i echo the above experience strongly.

However, it seems to me like there must be some middle ground, between sharing architecture "ideas" and full on code PRs.

I myself maintain various FOSS projects; and incoming PRs are marginally useful. generally i just close, and copy with claude following my own & project taste (crediting initial contributor).

▲pitchlatte an hour ago | parent | prev | next [-]

as someone who’s been working with MIDI for a hot minute, I’m really puzzled by your experience. either you have a very fussy controller or you received bad advice. most samplers like Kontakt and modellers like OrganTeq (very well regarded organ modeller) literally comes with a standalone version so that you don’t need to fire up a VST engine/DAW.

▲ValdikSS 3 hours ago | parent | prev | next [-]

>So maybe we should share OSS "templates" instead of full products, projects that are the barebones of a solution that you can clone, and have your LLM modify to suit your needs, on your platform of choice.

I've seen this idea: https://v-it.org/

▲timbert 2 hours ago | parent [-]

Thanks for sharing this link! It's a great idea, maybe I'll contribute as well :)

▲rabf 2 hours ago | parent | prev | next [-]

I think supplying the solid base on which to build along with instructions for a users LLM is very valuable just now.

▲larodi 3 hours ago | parent | prev | next [-]

readers here who have never tried this or does not understand what the person is talking bout - dare not speak. really, you have no idea, so please scroll further. those who understand will surely upvote. because _this_ is what savvy and pragmatic people (would and does) do in 2026 in order to get stuff done, perhaps they did otherwise back in 2012, but not anymore.

▲Conscat 3 hours ago | parent [-]

I think I principally agree with you, but did you need to write this in such a condescending manner?

▲augment_me 3 hours ago | parent | prev | next [-]

This is a take that is outdated if not now, then in 3 months.

These has been this ladder of copium that "oh human still has input here for X, oh shoot it doesn't with Claude 4.5, well human is still needed for Y, oh shoot...".

You are suggesting that you can provide context that the model can't on its own. This is a search problem, every iteration it gets better.

We are at the stage in our company where "/goal optimize this" gets better result than any biased human context engineering can provide.

▲zahlman 2 hours ago | parent | next [-]

> You are suggesting that you can provide context that the model can't on its own. This is a search problem, every iteration it gets better.

How exactly do you propose that the model should search my thoughts? Of course I can provide context that the model doesn't have. I can provide context that does not exist anywhere on the Internet (unless I got hacked).

> We are at the stage in our company where "/goal optimize this" gets better result than any biased human context engineering can provide.

First, a "this" must exist to optimize. Second, you have to be happy enough with it that optimization is the only problem.

▲shmolyneaux 3 hours ago | parent | prev [-]

I think it's reasonable that humans need to provide context for this, and situations like this, forever. The GP has a specific problem. The context of the problem exists in the real world, not something that the agent can search for. This isn't a "humans are still important" take. This is custom software for meeting the specific needs and preferences of an individual.

I fully agree with them that the easiest solution is to ask an agent for exactly what you want, and iterate from there. Evaluating and installing and learning off-the-shelf software is often more of a pain than doing it from scratch.

▲augment_me 2 hours ago | parent [-]

What prevents a solution where you physically hook up the needed components and then type "/goal make this keyboard extension work"

And the same library is built without the "seed" or baseline?

▲zahlman 2 hours ago | parent [-]

What does "work" mean? How should the agent know what UI you want? How should you know, until you try something out?

▲augment_me an hour ago | parent [-]

Answering here for both of your comments:

1) If the UX/UI is important for you, the agent can have access to your other work, or to your other conversation and build your tool with that in mind. This is just another search space question. If you have a documented, predictable taste, the models can efficiently produce what you need.

My uni supervisor/professor used to complain about my slop papers, then I started generating them in his writing style by just pulling down all of his papers and blogs, and he doesn't mind it now.

2) In general, the LLMs are great for steering opinion. There is good research on steering political opinion with LLMs. Your taste and opinions can be formed by the LLMs to like the solution that it provides you.

▲rewgs 37 minutes ago | parent | prev [-]

> I wanted to plug in my midi controller keyboard and just play a piano, so I can practice my choral music. Very basic. > > This turned into many hours over several days trying multiple DAWs, trying to get tehm to install, get them to run at all, get them to recognize my controller, get them to play a VST plugin...changing system settings, completely jacking up my sound, restoring it all again.

Hate to say it but this is genuinely just a skill issue.

Resorting to building a DAW tailored to your own lack of understanding because you can't be bothered to learn tools that are extremely widely used and well-documented is the most extreme case of yak-shaving I've ever seen.

▲IncreasePosts 14 minutes ago | parent [-]

OP got what they wanted in 3 hours.

Since this is a skills issue, could OP have learned everything they were lacking and got a more standard set up working in 3 hours?

Isn't needing to know how to manage and coordinate various software packages more yak shaving than just telling AI what you want to do? OP wanted to play piano and hear it on their computer with some effects, and that's what they got