Oh yes, the routers and gateways that most people have that are isp provided that may not actually have open VPN or wireguard support.

Those ones?

Also putting a VPN in someone else’s house so that all their Network traffic goes through your gateway is pretty damn extreme.

Nor will the VPN work on things like their TV or Roku or game console. You know the things that people typically sit down and watch media on…

Which doesn’t work for The grand majority of devices that would be used to watch said media.

Tvs game consoles rokus so on so forth typically don’t support VPN clients.

The Jonathan clients for these devices also typically don’t support alternative authentication methods which would allow you to put jellyfin behind a proxy and have the proxy exposed to the internet. Gating all access to jellyfin apis behind a primary authentication layer thus mitigating effectively all security vulnerabilities that are currently open.

deleted by creator

That’s an engineering culture problem. Not a PR problem.

Like I said plenty of products call this tab completion, and it’s context aware completion, or predictive completion. I used an overloaded term but I would have thought after my explanation you would have understood what I meant by this point. You’re continued explanation of classic tab completion is shows otherwise.

and way predates whatever VS Code may have been doing

Also I said Visual Studio, not VS Code. 🤦

Secondly, even if you want to move the goal post by talking about some specific implementation of ML based indexing, ML is not LLM.

I very specifically said that it was ML based, The word was indicates past tense. 🤦

“Modern versions of it are almost entirely LLM based.”

I don’t know how you managed to completely skip reading that last line?

Here we are though arguing over reading comprehension issues. Which honestly is pretty classic for the internet.

I mean, fundamentally, yeah.

But we live in a corporate controlled, corrupt, world and now of these larger companies can be trusted with this process.

Some smaller communities and platforms DO this right sometimes, as they build in house processu that respect privacy. But governments world wide are making this impossible through increasingly strict compliance requirements that actually increase data privacy risks and funnel these needs to 3rd party services who just lie about what they do with the data.

===========

I’m not kidding when I say this is a REAL BIG PROBLEM.

bot based traffic and astroturfing will supplement and replace human communication on platforms like Lemmy. Driving the narrative and how we engage to the whims of a few rich people. Bots are relatively cheap, and easy to deploy at scale across many platforms.

There will be no open corner of the internet safe from manipulation and forced division. More people will be forced into walled gardens from corps that implement human verification, as they are the only ones with the resources to do something (While also being the source of the problem, see how that works?)

How do you carve out spaces that are protected from that? Well, you need to determine who’s a bot, and who’s and actual person.

But we can’t do that, so the alternative is we are ran over by bots and astroturfing till we’re at each other’s throats like good culture war puppets.

The future is bleak…

It’s largely considered ineffective these days. Detecting elements that don’t affect layout is trivial, or elements that are occluded, transparent…etc

Capchas are one of the best options. But even then, LLM users bypass those relatively easily, and LLM users are one of the biggest risk areas for astroturfing.

Apparently meaning from usage cannot be inferred here? Or you’re just being intentionally obtuse?

• predictive completion
• context based completion

A not insignificant number of products literally just call it tab completion these days, because tab completion in many products & IDEa is by default predictive completion, which is ML based. And these days, LLM based.

I used the wrong term, but I guess y’all were unable to infer meaning from usage?

Auto tab or w/e it’s called (Some products literally call it tab completion). Visual Studio was doing it around 2018 IIRC, it was ML based, always has been. Modern versions of it are almost entirely LLM based.

The best

Literally every game that’s made today is using AI as part of the development process.

Damn near every Dev has tab completion on in their IDE. Which is AI based.

==========

Edit:

I used a term, but I guess y’all were unable to infer meaning from usage?

Auto tab or w/e it’s called (Some products literally call it tab completion). Visual Studio was doing it around 2018 IIRC, it’s ML based, always has been. Modern versions of it are almost entirely LLM based

Ironically, I am one of these people building get up action CI pipelines that automate large multi-service multi-team deployments across different cloud vendors and regions.

It’s all a bit of a pain in the ass and. Actions don’t really provide many of the nice controls and safety guarantees we would want.

However, GitHub actions is relatively straightforward to implement and relatively easy to operate.

It’s just very easy for you to foot gun

Training is constant. None of these models by any of these providers are static. You’ll notice that they are releasing new models and new model versions regularly.

This means that training is happening constantly. It never stops. There’s always new shit being trained.

You’re talking to non -tech nerds about something that usually only tech nerds are familiar with.

Just like on Reddit you’re going to get downvoted because people don’t understand. Lemmy is effectively the same in that regard.

Just having data is cheap. Actually serving that data up in a meaningful way is expensive as fuck.

Yeah if that’s the case they really kind of fuck themselves by not being strategic in how they announce this.

Announcing it first so you get all the heat before the date is even released is a great way to sabotage your mission.

Biggest problem for Lemmy and similar applications is scalability and controls and detection for bots.

The compute costs to operate instances are astronomical compared to the actual user load they receive.

That’s a bit of a side problem compared to bots though. Bots are a real problem that services like this are not equipped to handle.

They are already arriving to some degree.

The difference being is that Lemmy and other similar services have zero controls or ability to handle bots or bop traffic if those bots were bots from 2014.

Not bots from today.

It’s a bit of a problem and honestly with increasing bot traffic across the internet and fedaverse being extremely vulnerable to it It’s absolutely bat shit insane, but I don’t see any other option than somehow having some form of human verification.

It’s a problem

I’m guessing I must have missed something here when I made that comment. I visited the link in the body of the OP not once, or twice, but three times to verify I wasn’t losing my mind. Even went into reading the readme, some issues…etc to verify.

I’m now realizing that in my Lemmy client the link in the body is more obvious to click on than the actual article itself.

Did you go to the repo before running your mouth? It’s awesome-selfhosted data.

What AI slop?

Edit:

I’m guessing I must have missed something here when I made that comment. I visited the link in the body of the OP not once, or twice, but three times to verify I wasn’t losing my mind. Even went into reading the readme, some issues…etc to verify.

I’m now realizing that in my Lemmy client the link in the body is more obvious to click on than the actual article itself.