m0wer

You missed the point. The IP protocol doesn't scale! Ethernet MTU is just 1500 bytes how are we ever going to transfer a movie over the Internet!

Ethernet does not need to carry the whole movie in one packet. If it does the job of delivering the MTU to the host on the other side of the cable, it's good. Websockets can be figured out somewhere else. The IP stack is not shit because each layer does just one thing, it's good because of that.

But hey, time will tell.

Not sure. IMO the best thing that could happen for the next generation is to be born in a Bitcoin standard were politicians don't control money, people are incentivized to save, and the world does not need to use housing and stocks as a way of saving and protecting against inflation.

Technological progress makes the world deflationary. Your money should be able to buy more every time as we improve the productive efficiency of everything. And for poor countries, the best thing they could get is a censor resistant and value preserving tool.

Even if there was a tail emission, newer generations wouldn't have the capital needed for mining rigs. That's not just something unique to this case, same happens with stocks, real state or any other investment asset.

Isn't being able of anonymously send you something of value that no one can take away from you a pretty big use case?

A third of the world is unbanked. A permissionless monetary system makes a huge difference for those.

When I was still very skeptical about Bitcoin, I met a guy in Turkey who was from a very poor African country and was just studying there. His father would buy Bitcoin in their home country with the local currency (P2P) and send it to his son, that would then convert it also P2P for Turkish Liras. They could do this securely an within minutes. The alternative was using Western Union and paying taxes in both countries, which in total added up to ~50% of the sent amount.

It's great not needing Bitcoin, as it is great not needing Tor. But that doesn't mean there's no use case for them.

Blockchains don't scale. But that's a feature, not a bug.

Great protocols are built in layers.

You have decentralized instant settlement for an average of 0.005% even for micropayments with the Lightning Network (another protocol built on top of Bitcoin). That's orders of magnitude away from the settlement time and resilience of the current payment networks.

Would it change your view if they mined instead of buying?

If you were to create a decentralized and limited supply currency, how would you distribute it so that it's “fair”?

Sounds a bit like if the world was running only on proprietary software created by Microsoft and you criticized the move to open source because that would enrich Linus Torvalds and other code creators/early adopters.

Are people better off by continuing to use centralized broken software that they have to pay a subscription for (inflation) than if they did a lump sum buy of a GNU/Linux distro copy from a random guy and become liberated for the rest of their life?

The “cool” thing about embeddings is that you can use a generic one that can even support multiple languages. Just with the plot/description of the movie you could see “similar” ones in different aspects such as movies that also talk about dogs or more complex relations. Here is a short post about how it works: https://blog.sgn.space/posts/embeddings_based_recommendation...

Probably what you want is not an LLM but just the embeddings for clustering. It's much lighter and would work well with new material as well.

I've tested it out for filtering RSS feeds and has worked pretty well [1].

[1] https://github.com/m0wer/rssfilter

Always has been.

About a year ago I created rssfilter to help me out catching up with my RSS feeds.

I was getting too many new posts per day from HN and other sites and started not enjoying it as much. I wanted to have 50% of the posts but missing out as little as possible about the topics I'm most interested in.

After a year using it, I'm pretty satisfied about the results and other users are too.

How rssfilter works is by being a proxy to your existing RSS subscriptions. It replaces the article links to a redirect that stores what articles you open. This information is what's used for the recommendations.

The embeddings of your read articles title and description are clustered. And from then on, you will only get 50% of the articles from all feeds, mostly what's closest to your areas of interest but also some percentage of unfiltered posts to allow for discovery.

It can be self hosted and there is also a live version that I host. No registry, the user ID is just a random UUID.

You can also use the API (FastAPI btw) directly: https://rssfilter.sgn.space/api/docs

Let me know if you like it!

TabbyML! Autocompletion like GitHub Copilot, using qwen-2.5-coder 7B.

Faster and better experience than DeepSeek Coder V2 Lite (which was already impressive). I've been using it for the last few days with Tabby + Ollama.

feeder + rssfilter.sgn.space

not working...

Pretty accurate predictions about Linux servers, startups, and remote work.

Was submitted already 12 years ago but got no traction: https://news.ycombinator.com/item?id=4202529