A significant contribution to that is residential power effectively subsidizes data centers for AI.

Felons doing felon things. Have we no self-respect?

I’ll bet you $100 what side Congress will take in the matter, and it will be on the side of money.

I look forward to my first $400 power bill.

You don’t get to order people who don’t work for you.

We’ve got representatives attacking the postal service for not being highly profitable. We aren’t at square one of investing in the public good even when it pays in the long run.

Costs are borne by all, but profits only for the few.

Then, he is a fool. LLM technology has no fence around it. You can download and run one on your own hardware. The only reason a person would use their service is convenience access to a larger model.

I appreciate a simple piece of software that does exactly what it’s supposed to do.

This is the way to do it. If you’re paying for your whole PhD, you’re doing it wrong.

He does know how to get down on his knees.

We need to stop saying “American Dream.” It’s really beating a dead horse at this point. It feels like such a fictional concept to me.

My friends and I talked way back in school about how further engineering education was negatively correlated (not exactly: see comment) with pay after a bachelors and was statistically a terrible deal.

EDIT: That’s not to say it’s worthless! But it ain’t worth what they’re charging. There isn’t actually a negative correlation in the strict sense but rather there isn’t clearly a premium for the degree in all markets. You can be taking a straight up financial loss. The original statement was inaccurate, but that’s historically what we told each other.

I will continue to enjoy my incredibly straightforward and to the point Linux desktop that’s somehow gained a new AI-free feature by doing nothing.

I’m concerned about the large amount of low quality, vaporware/crypto applications built on IPFS which is the same core technology used here. It’s concerning how many clicks it takes to get technical specs for the underlying work, like libp2p for the network layer, which itself espouses only vague ideas on its main website that seems to focus a lot more on presentation than technical merit. Even the GitHub admits that the spec that most of these apps are relying upon is, well, unspecified.

Your project source downloads and runs an executable. That’s a little bit SUS; it would be much better if you compiled/built this core code as part of your build process, else, it’s not much in the way of source code, no? But, it works. It seems to delegate just fine, and few understand how to actually talk IPFS directly. But, this is the most important part!

I think the biggest tell that IPFS borders on vaporware is that there’s very little discussion about concrete specifications and the main problem faced by all DHTs: how you get your data to actually stay hosted on the network over time. These ideas are not new, and you may be better served building your app on technology that has spent vastly more time understanding the fundamental problems.

https://specs.ipfs.tech/

This is how you write a spec without actually writing a spec. And I’ve written a lot of specs.

https://geti2p.net/spec

This is how you write a spec. Excruciating detail of what actually gets sent over the wire at different levels of the design starting from the very bottom.

Anyway, just my 2c. It’s cool you’ve got functionality at this level and that’s commendable, but I feel it’s built on shoddy foundation of an immature technology. At least it should be easy to migrate to something else in the future as the distributed technology is offload to a separate binary anyway.

Note: Various edits for clarification and to ensure I focus on the code and not the human.