Hacker News Reader

Comment by cortesoft | Hacker News Reader

This seems like solving the problem at the wrong layer? The issue isn’t the actual network connection between people, it is the content. You could easily create your own forum or something and only include people you trust. You don’t need an entirely separate internet.

View on HN

>The issue isn’t the actual network connection between people, it is the content.

Everyone serving a website is being ddos by AI agents right now.

A local mesh network is one way to make sure that no one with a terabit network can index you.

Then firewall traffic that doesn't come from your local ISP blocks or authenticated users.

You could set up two way TLS with client certificates

That isn't good enough and could be DDoS'd as well.

There's a dead comment here saying that OpenAI doesn't just DDoS the internet because it can. That's true. Any supposed DDoS is a side effect or incompetent scraping, and won't affect anything they can't scrape. Not sure why it's dead — it's important to realise this.

Edit: oh, it's probably dead because of the username

Could you just geoblock the USA? Is most/all AI agent scraping from there?

Geoblocking is nontrivial

If it's a mesh network your peers are tech savvy enough for a mesh vpn like wireguard, which also doesn't get ai-ddos'ed.

Make it private invite only behind Cloudflare

Oh yes, just let Cloudflare solve all of our problems. Fuck this. We shouldn't have to rely on yet another huge company to fix this for us.

>The issue isn’t the actual network connection between people, it is the content.

>>Everyone serving a website is being ddos by AI agents right now.

You’re missing the point, the point is that while mesh networks solve a problem, it’s not required to solve the problem ”I’m tired of the Internet” or ”I’m being indexed”. You can build your own network on top of the Internet with zero new hardware required, with something like wireguard, i2p or whatever.

I don't think they ever claimed that it was the only way to solve the problem, just that they're doing it and it works for them.

I was able to block them on my silly hobby domains. Most of them were already blocked to begin with from blocking other shenanigans over the years. Even something as simple as blocking anyone that does not support HTTP/2.0 takes out most bots. Adding basic-auth also stops most of what gets through. Blocking TCP-SYN with strange MSS values cuts out many before they can even touch the web daemon.

Even if it was a "network connection" issue creating an overlay network on top of the Internet (with VPN tunnels and mesh routing, for example) would yield wildly better bandwidth and latency characteristics.

You can still make that overlay network geofenced and vetted. Heck, running it over a local ISP's last mile would probably yield wonderful latency.

We need vetted webrings on the existing Internet, not a new Internet.

Reading this back and forth so far I think you’re spot on… which leads to this open question, wheres the consolidated stack that makes this accessible?

Also I think the name vetted webrings or just the vetted web is simple enough to be a movement.

As in the vetted web movement.

… gotta start somewhere.

But it’s a whole lot less fun and educational than building your own infrastructure. Sometimes the journey is more important than the destination.

> We need vetted webrings on the existing Internet, not a new Internet.

How do you think will this work, when LLM accelerates the breakdown of trust and common epistemics?

There's only so much you can do to detect and block content that's AI generated. At the end of the day, the content starts with the people creating it.

Jumping to an invite only network isn't the most ridiculous idea imo.

What parent means is that you can with no problem build over the classic tcp/ip.

The best solution for dealing with AI content slop flooding your eyeballs is to hang out in places small enough to be a community -- like a local area mesh network.

AI slop thrives in anonymity. In a community that's developed its own established norms and people who know each other, AI content trying to be passed off as genuine stands out like a sore thumb and is easily eradicated before it gets a chance to take root.

It doesn't have to be invite-only, per se, but it needs to have its own flavor that newcomers can adapt to, and AI slop doesn't.

You can still find the essence of community on the traditional internet in places like invite-only discords, smaller mastodon instances, traditional forums, and spaces similar to Lobsters and Tildes.

...and not on Hacker News. Too many pseudo-anonymous jerks, too many throwaways, too much faith placed in gamified moderation tools.

Potentially, but those areas are also more and more getting leveraged to further identify and profile people for targeting - see the latest Discord scandal for example.

This is technically correct but there is no harm in them setting up a mesh in the event that the internet goes down assuming they have their own backup power. The USA and EU are both extremely vulnerable to power grid overload, cyber-attack, physical attack choke-points leading to black-starts, EMP, GRB and much more. I think it's good on them for the learning exercise and hopefully they add to existing documentation.

If the power grid overloads, I am not sure that your mesh network will survive either?

It runs off solar.

I'd like a semi-anonymous private network. Something like: I go to local post office and purchase a sealed token. I use the token to generate a reusable “verified human credential” with limited reuses. The credential allows me to connect to the private network.

I learned about a cryptographic interaction that can support that recently (and have spent a lot of time focusing on the idea as a means of procrastination).

I don't use Kagi but the context was their Privacy Pass thingie https://blog.kagi.com/kagi-privacy-pass

It works similarly to what you'd like: they sign sealed tokens you provide. Later, you can unseal a token and use it without invalidating the signature. It is mathematically too difficult for a classical computer to link the sealed and unsealed token.

You’re going to end up running down the same merry path that DRM companies do - and you can’t patch the wetware layer. Inevitably thousands of ‘human tokens’ will end up in the hands of actual humans working in call centres with 300 phones in front of them.

https://www.scmp.com/news/people-culture/trending-china/arti...

What happens when Comcast is told to cut off access to "blue" cities / states?

> This seems like solving the problem at the wrong layer? The issue isn’t the actual network connection between people, it is the content.

Classic HN. Focus on the tech to avoid looking at the problem.

Perhaps, but it also, by default, excludes that entire class of authentication problems that are only manifested in a non-local network.

I love the idea.

It's also interesting in that a local mesh doesn't necessarily need to operate using the TCP/IP/HTTP stack that has been compromised at every layer by advertising and privacy intrusions.

You’re probably getting downvoted because what you said about TCP/IP/HTTP doesn’t make sense.

You're right. I didn't think that through. The stack doesn't imply that a local network is somehow exposed to those concerns.

Reply on news.ycombinator.com