They do have class compliant USB audio, but just for the main audio mix, not multitracked. I actually end up using that a lot more than overbridge.
But in fact for me the problem is that when you go for a non standard implementation, you may gain something but you lose compatibility. If you then have to do a specific driver you need to update that every time a new OS version comes out, win, mac and linux. Ok, i’m jokin, not linux. 
Is it worth? and how long will you do it?
I bought long time ago a firewire audio card, cause “firewire is better”. It became unusable after few years, because there were no drivers for the new PCs. What a waste of good hardware (btw was a Tascam Fw1804)
1 Like
I imagine Elektron tried to control both platform support and end-user performance experience. Like building x-platform UI libraries allows for (close to) a single UI design and codebase independent of platform, I would imagine OB allows Elektron to specify a more performant interface regardless of host OS.
You’d think USB class compliant would do the same, but from recollection, Windows default audio subsystem is a pile of shit (it’s been 15 years since I used Windows regularly for music so you can school me of my wrongs if you think I need it). Elektron probably wanted to have more control over audio streaming performance than they could have got relying on standards.
(The USB consortium have an increasingly hostile attitude towards end users; the specs are designed to cater more to efficient/cheap manufacturing than they are to a good end user experience)
3 Likes
Yes, but then you are solving a problem outside of your domain. Is that worth it, given all the drawbacks that come with it? Being plug’n’play it’s a great thing, and this implies class compliancy.
The point is that “AI” hype and ML are two separate things.
The former is overbroadly hyped as an all-consuming force, not a tool for enhancing productivity.
The latter is best used as a targeted tool, and useful. We’ve all used and had ML assistance directly or indirectly, it fades into the background.
This delirium-tinged futurology is unhelpful and more about how technology makes people feel to listen to TEDx talk than is based on actual utility.
Just remember that similar or the same people said the same thing about cryptocurrency.
Focus on actual utility, not the hype cycle. ML existed before LLMs. The hype is exclusively over LLMs, not usable ML.
Yes, nothing has materially changed there.
Even beyond when the rubber meets road, “simplicity” is complicated, knowing the least code you can write, the most common failures, the most viable paths before you begin to actually sketch anything out.
I know we’re getting fairly far away from @Sternenlicht’s original post (if it’s even referring to them!) into broader discussion trends, technology is magic to some people, and the less a person knows surface-level the more we get into ego, Cassandra complexes and overbroad statements that gloss over implementation, precedent, and are too entwined with press releases and other unpaid evangelists who love the feeling of “newness” that trend-following gives.
4 Likes
For streaming audio, from an audio tools manufacturer - yes. Well, Elektron thought so.
Sure. I’ll take reliable audio over plug&play.
My last few audio I/o cards all had dedicated drivers, even for macOS which has a good audio subsystem.
1 Like
I’m struggling to make LLMs useful for my day to day. I don’t struggle to write letters or code and feel like I spend time learning how to write prompts that generate code I could have written myself. I like your reasoned take on the “threat”.
There’s one point in this “fear-mongering” that’s worth paying attention to, and which is getting lost (in this thread and elsewhere). The current tools are mainly closed and controlled by heavily capitalised private funds. OpenAI are everything but open. Meta may be publicly traded but good luck getting around Zuck’s 51% voting rights. Microsoft only embrace open-source when they’ve worked out a way to enclose and capture value from the community. The original Luddites were not reacting to “technology” in itself, but to the vastly unbalanced capital distribution between tech’s owners and users/workers, and the ensuing ability to earn income. The fear-mongering over lost jobs is as over-hyped as the tech, but the core problem of wealth distribution remains and should be talked about more.
(yes, I know there are open LLMs, but they’re hard to run for lay people. I’m a career coder of 20+ years and spent my teens fooling about with tech as much as music… and I would struggle to make use of the open LLMs without serious study).
4 Likes
Things have improved over the last few years WASAPI is very close to ASIO in terms of latency these days.
This thread was never about AI. That’s a hijack. Maybe we need a moderator intervention.
1 Like
It’s heartening to hear there have been improvements.
However, I’m a snarky nerd this afternoon and when I read this I thought “well, ok… so Windows has an updated thing that’s almost as good as the shit it had in the early '00s - doesn’t sound like much of an improvement”. I’m having a bad day so don’t take this seriously tho’.
2 Likes
I for one welcome our AI overlords.
Haha, tbf that’s totally valid! WASAPI is nice in that many apps can take advantage of the lower latency for free as it defaults to run unexclusively, it’s made web audio latency better on Windows for sure - and late is better than never 
1 Like
It’s fine for simple things, but even they require cross-checking and auditing maths.
Speaking from anecdote because pertinent but not meant as any authoritative viewpoint-
I know someone involved in architecting widely used piece of software, since AI bumps stock price it will be incorporated into the product but… the maths just aren’t serviceable in comparison. The main product’s is used for all sorts of vital calculations, but this immediately breaks down when you substitute a valid, testable engine with something that at base, is not about mathematics at its fundamentals, and requires constant human intervention (hence the “Clever Hans” puppetry)
I’ve used it to calculate the amount of wood needed for a particular project, I use it for short scripts, I’ve tried to use it for abstract and advanced topics (music theory) and attempted to use it for more complicated questions about coding architecture and I’m better off using my brain.
Yerp, whether it’s crypto, AI, or whatever the next stock hype cycle gets into, it’ll always have excuses made by the people who want to get in on the ground floor, and it’ll always be driven in a direction that exacerbates the incredible growing disparity.
“AI” hype is all about complicated scenarios being “solved” (but not really) and people with less real-world experience believing they know better than people who use ML and “AI” tools in their daily lives.
Simplicity of complex systems is the same topic, from a different vantage point.
2 Likes
Your hidden agenda was to group all the software devs in a thread so to “racially clean” the other threads (being software devs a clear race). Your AI overlords asked to do that, devising this perfectly crafted title for catching our attention. I know cause I asked ChatGPT.
2 Likes
We’re onto you, Deep Thought!
First, the Genuine People Personality. Next…
1 Like
Well … I suppose, in the end, everything is related to everything else … but really, the threat (or other wise) of AI has very little to do with people requesting firmware updates on sequencers because “it should be simple to code”
And all we get is something that looks almost, but not quite, entirely unlike code.
2 Likes
Meanwhile in a parallel universe…
Elektron guys on the Windows forum open a thread “usb multichannel audio latency sucks”, they write: “we can’t currently use it as it is now, can you please fix it? We could it ourselves, should be easy to code”. And the answers go like: “what do you know about coding? It’s always harder than that! Nobody is really interested in such niche functionality, go back nerds. Do it yourselves, then brainiacs!” Then on that thread they start to talk about AI.
3 Likes
“all this has happened before, all this will happen again”
So say we all…
1 Like
They both come from similar vantage points and psychology around them, hence the initial reference regarding both public understanding of information science, the economics of it, and how capital would also love to devalue the art alongside.
AI isn’t the point of the thread but is relevant to discussion of how cargo cult-ish public conception of technology is. Historical cargo cults are also not the direct topic, but are the topic as well 