Sven Slootweg (soft-deprecated) @joepie91@pixie.town

@jonah I do sometimes get the feeling that this is the intention of some projects... and that people are just looking for a way to confirm their prejudices.

More specifically, the 'red flag' to me is when a project is asking for an amount of donations that's so large, that it's obvious they haven't really tried to reduce operational costs, and instead went for the nearest most convenient or familiar option, usually one that's explicitly designed for corporate use. Think stuff like building everything around AWS.

Even if they don't *intend* to monetize things later (but it's likely that they're not very opposed to it, because it often signals a corporate background), with that kind of operational cost, they will eventually *have* to monetize things simply because the operational cost is not sustainable.

It kind of feels like the most reliable predictor for "does someone intend to eventually monetize this donation-based thing" is the amount and scale of donations that they are asking for, with an exception for projects with a very large scope

@admin Huh, odd, I feel like it used to be a list of just the city names without the account address. Perhaps I picked it off a different list?

@admin Two bits of feedback:

- It seems harder to navigate the list of cities now, there's a lot of spurious information in the city list that isn't the city name
- Not a new issue, but <<'s Hertogenbosch>> is being auto-capitalized incorrectly (see attached image)

@mitu @davidgerard All this is skating right past the most crucial question here: why would they remove this to begin with? That is a deliberate action, it doesn't just magically happen, so there is an intention of some sort behind it.

@starfire It is indeed closely tied to Chromium (or more specifically CEF), that part is correct, but you can just have one system-wide installation of Electron the same way you do for Node, and run `electron path/to/application` and that's it. This is also how Electron worked from the start, the whole 'bundle a runtime with your app' only came later in Electron's life.

@starfire What I really hate about this particular myth is that it basically only exists because of commercial incentives; it's cheaper for a company to ship the runtime with the app (and have a consistent environment) than to use a system installation of Electron (and have to support users in different environments).

The result is that, because corporations drive almost all software development nowadays, people have started to believe that it is a *requirement* to do this with Electron, because obviously the tools have started focusing on this approach as well because that's what companies want.

And so a bunch of misinformation spreads because "Tauri doesn't need this!" even though the moment Tauri were to gain widespread adoption (if at all), companies would immediately construct a very similar setup because they just fundamentally don't care about the cost to the user, only about their own support cost, and the tradeoffs would be the same there.

So now there's a bunch of misinformation and a lot of wasted resources and work being replicated between projects and unnecessary tech stack migrations and now everything is worse for everyone involved because a bunch of tech companies wanted to skimp on support cost.

@starfire That last point is actually false; it's comparing an Electron application with bundled runtime to a Tauri application that uses system libraries, but it is entirely possible to do the same thing with Electron and also only ship your application code.

(This is a frustratingly persistent myth...)

@SteffoSpieler @finn A small (but important) correction: there's a lot of talk of "open-source AI models" but the last time I evaluated this, a couple of months ago, exactly zero of the models I looked at from the LLM/transformer corner of the industry were *actually* open-source, despite claims to the contrary.

Often it really just amounted to "you can download the pretrained model" and it was impossible to replicate (or even modify) the training process because one or more components of that process were completely missing in what was released, usually also with little or no information about where the training data came from.

As far as I can tell, when it comes to transformer/LLM/generative/etc. stuff, "open-source models" simply do not exist as a category outside of a few academic experiments that produced useless results.

@starfire Hmm, last time I looked at it it definitely didn't do that (unless you treat it like any other Rust library and use it via Neon or whatever).

Do you have a link?

(That having been said, from my last evaluation it was basically equivalent to Electron so while it's *technically* another option, the list of tradeoffs would look almost identical)