>* Is it the YouTube algorithm's fault? Or viewer preferences? I think YouTube would be happy to recommend the videos to more people if more people watched them.*

I always got garbage even if it suggested things I wanted to watch too.

The best way is to disable suggestions completely and then just make a note of your favourite channels. That way you get a completely blank landing page and are forced to search out exactly what you want every time.

You can disable suggestions and still subscribe to channels, I'd argue that's the easiest path while still keeping what you're actually interested in. If you have 200+ subscribed channels over the years it's less good, I'd just unscribe if you're not interested in it.

You can still choose not to buy the “authentic“ good can’t you?

I can, but what annoys me is people saying "Well that technically isn't champagne because it isn't from the right region of the right country" and acting like being able to quote obscure regulations makes them "technically correct" in any context outside of a courtroom.

Yes! Always run some numbers in your head. People will sometimes pick percentages or numbers in an attempt to bolster their arguments. Don’t buy it, work it out for yourself.

Those are just products who cares?

I think the GP is alluding to understanding the fundamental way a thing works.

I think you need to ask what you actually want to do with the AI.

If you want to be a researcher and come out with the next breakthrough, get ready to go back to school and learn some math.

If you just need to learn how to use it well and build things with it, then you probably just need to have a high level understanding.

Same as programming. I’d bet most programmers have no idea about the physics that makes computers work.

> I think you need to ask what you actually want to do with the AI.

What about improving the efficiency of token consumption, etc., basically opportunities for improving cost/performance?

I keep thinking there has to be a better way to share context with models than dumping entire gigantic skill files of raw text or otherwise into them - I'm betting there's a bunch of low-hanging fruit there.

There may be some low hanging fruit, but they're not available to people without deep understanding of how the math works. Well paid people already spend a lot of time thinking about this.

i am not sure acctually of the math is acctually that complicated/important. the math around neural networks is calculus/chain rule etc and for model comparison/validation one needs statistics. the required math for e.g. understand transformers is quite accessible.

You missed the third and most important reason to learn: fun.

Which sums up HN these days.

I used Linux for a long time. I still prefer it. But I can’t justify the extra work. Last time I tried to move back to Linux, I spent far too long admiring the machine. This was only 5 years ago.

AI agents are incredibly useful in this regard. Omarchy even releases some skills, so anything you want to configure is just a matter of asking the agent to do it.

Of course, it would've been better not to need it in the first place. However the experience is much better now (and Omarchy is great!)

I shouldn’t need an agent to configure it.

Is this the new “just compile it from source”?

You do not in fact need an agent. Without it, you'll just need more time to inform/educate yourself and do it yourself.

Agents in this case are just accelerators.

Agreed. I've setup many VPS servers. I've configured both Apache and NGINX by hand. It's the sort of thing that I only do sometimes, therefore I have to look up docs each time.

This year, I wanted to run an experiment that was extremely low stakes. I had CC ssh into a VPS and do everything for me. It worked.

Did it have any security gaps? I don't know, I didn't audit. It didn't matter. It was a personal project that had no personal data to risk and the project lasted ~1month (I'd actually audit if I was keeping it). I saw no signs that my server became a bot, fwiw.

I've had agents figure out git repos that were interesting to me but not worth putting in effort on my own. Not things I depend on, just things that I otherwise wouldn't have tested. It's too bad the subsidized pricing is coming to an end. There's a lot that I wouldn't have bothered doing myself that was cheap to have an agent do that won't be worth it again. I'll try out a fraction of the stuff by hand like before and that's fine, but it was a fun ~six months.

I truly feel old when we are rehashing the same thing that happened when TPB, Limewire etc were taking off.

Down to the arguments being made why this is “theft”. I also can’t believe that the arguments by the general pop have shifted to support multi billion dollar companies and their narrative.

Just because people do buy the product the legitimate way, and provide the creators with revenue, doesn't negate that /you/ are committing a theft.

A loss of a sale, is still a loss of a sale. It's vastly different to /not/ selling something.

There is nothing to negate because there was never a theft in the first place. Consult the laws of your country before making wild assertions and ask a lawyer if you cannot understand them yourself.

It’s like saying the death penalty isn’t murder because the law says it’s not.

I guess in your mind hitler isn’t a mass murderer because what he was doing was technically legal under his regime. To each their own if that’s how you want to logically frame your moral logic.

Personally imo, if I create something and that something is MY work of art and what I put my own blood sweat and tears into, and I want it to be inconvenient to watch because that’s my prerogative, that doesn’t give you the fucking right to watch it. I don’t want someone like you watching it.

Wow. Bringing up murder and hitler when talking about piracy. Some parts of the internet are truly eternal.

None of this is new. None of your ideas are unique. Stop attempting to draw moral equivalency to theft or anything else.

Back in the glory days of piracy, profits from entertainment kept going up? You know why? Those pirates were almost certainly not going to pay in the first place. No one lost money. It was university students, high schoolers, plain old poor people that were pirating.

Making it hard to watch your movie didn’t raise revenue, we all just did something else. Because we had no money.

Logic and human nature becomes more apparent when you look at the extremes.

Flaws in logic and the darknesses of both become readily apparent because the paradoxes are exposed at the extremes.

Seems like the opposite to me. We don't like theft because the victim loses something. If stealing a car spawned a new car for you while the "victim" kept his, would it be a problem? So I don't really get the murder analogy. It's more like if the government had defined some act as murder even though nobody died.

Microsoft has built some good hardware over the years. The problem with this is that it runs windows. The hardware is probably nice.

Is this not just a tooling problem?

As someone who supports Atlassian products; Atlassian is a tooling problem.

Yeah, having two tools when you only need one is a problem. Like one is always going to be full phat and the other will ride the back of the bus.