The problem is, the places where it could be the most effective are places where we already have processes in place, like the medical field and engineering.
I’m sure LLMs are helping sift through ideas in those fields, but they don’t reduce the amount of work that needs to be done by actual people to produce actual results; they just help select the best potential new idea.
Most stuff in life is already solved, and so AI can’t contribute there.
And yet… it’s orders of magnitude better than anything that came before it, and nobody’s found a better way to do it.
I’d like to know which non self-hosted service they’re recommending that can serve files to my classic Mac collection for less than my $200 mini pc.
Or which VPS they recommend that would have taught my kids how to set up services and repair/select hardware.
Unless you’re purely a technology consumer, starting off with self-hosting is the only way to learn all the skills you need to determine what needs to be self-hosted, what needs a VPS, and what needs to sit on someone else’s technology stack.
Realistically, self-hosting has cost me around $150/year for the past 20 years. But that money has provided services and experience that either isn’t available or is MUCH more expensive to do any other way.
Generally the costs in self hosting are in paying for lessons learned as you make mistakes and have to pivot (hardware selection, software gaffes, etc.) and time. But those are costs that are easily recouped through personal development over time.
Other Chinese models aren’t on the list though, like Qwen or the ones from Xaiomi.
Is there reasoning posted anywhere?
LLMs generally don’t create anything new; they just amplify and sometimes clarify the inputs.
An infinite number of. With models, the concept of backdoors becomes a bit fuzzy.
It’s often a bot.
I mean both, for different purposes. They are definitely very different.
It’s worse than that even… frontier models now have guardrails restrictive enough that I can’t use them for legitimate work. And the ones that DO work have token prices at exorbitant rates, and chew through them like there’s no tomorrow.
Meanwhile, I can download Qwen3 and run it on my own hardware, and it’s about as good as the last generation US model without the annoying guardrails.
Hmm… which model am I going to choose to use?
And this is why I block all ad scripts. Sure, I might want to support the website, but I’m not going to run an unvetted third party’s scripts. Display passive ads and I’ll give them my views. Curate them appropriately and I may even click through.
The big issue is when older communication standards stop being supported and you lose access to 3G, 802.11n, TLS1.2, etc. at that point, the device becomes an offline-only device.
That said, I have an iPad 2 I picked up in 2011 that still functions perfectly except for web browsing. Works great as a jailbroken emulator, an eBook reader, running all the same offline apps I used in 2011, and the original battery still lasts for 8 hours on a charge.
Seems like it should theoretically be possible to have AI subscribe to all the news feeds (AP, Reuters, NewsWire, public reporting databases, legal and police feeds, websites of all registered non-profits, etc.) and generate hyper-localized page content linked to an ad network.
As long as you pull in more in ad revenue than you pay in tokens and infrastructure costs, it’s a win — once someone sets up such a system, they just need to fix things when they break; most of the process would be automated and self-healing.
I’m surprised we haven’t seen more attempts at this. Would it be slop? Definitely. But it would be profitable slop.
Well, it was inevitable.
And it’s between people who we already know are morally compromised.
I really wonder why they’re being called GPUs and not NPUs — they aren’t primarily designed to process graphics (although they can).