oops..

like

["e", "999", "wss://relay.io", "root"]

with 999 being event ID and the root marker meaning that's the one that started this. Something like that.

Also NIP17 has a subject-type tag that people can just change randomly, but for NIP04++ email you'd want to make it so the email subject tag holds firm.

So subsequent messages like

Ah yeah, mostly the same, but one less event.

NIP17 in a group of 5 people is 5 events

NIP04++ is 4 events, since you don't need a separate one encrypted for yourself (they're all encrypted to you).

And there's enough to work with in the exposed metadata to make these kinds of split-view attacks not worth trying to pull off.

For group email, no, not really. Normal email has several baked-in protections that make this kind of context injection / split-view tomfoolery a lot harder to pull off. You have to go back decades to the very early days of email before you get to a situation like we have now with NIP-17 groups.

There are some workarounds, most having to do with every client trying to keep a spying eye on every other client in some George Orwell 1984 JOSN dystopia, but they all end up in this uncanny valley between NIP04 and MLS, too hard for people to understand and not enough payback for being too hard to understand.

You could do nostr email with NIP04++ DMs though.

It's just online betting, like 90%.

Yup. NIP17 I just don't get. It works well for one situation: two humans who trust each other and each have their own dedicated relay. Which, I grant, that is a real situation, and it's a novel solution there, but that's it.

Two humans sharing the same relay, you can easily work out who is talking to who, so may as well just have the NIP04 exposed metadata so you're not fooling each other. NIP17 groups is a non-starter for me. And then the giftwrap spam/client-side processing tax on top of it all. I just dunno.

Yeah Nip04 with NIP44 encryption is the best thing out there. So obvious to anyone what it does and does not do.

You want to private, use a different key. 🤯

You're one of the screamers, so good on you, it's working.

One cartoonish (but helpful) way to start is to picture an advanced AI looking at all this bitcoin compute (all this massive processing thrown at factoring random numbers) and pulling it's hair out and screaming JUST GIVE IT ALL TO ME.

As bitcoin miners switch from ASICs to GPUs to earn instead from model compute, what they are doing in a way is listening to this JUST GIVE IT TO ME scream.

Eventually most of the bigger miners will move their farms to GPUs, even though it's a total refit, because why mine bitcoin at scale and earn x when you can provide model compute at scale and earn x + 1. Left mining Bitcoin will be smaller firms for whom the revised difficulty adjustment still makes it worth it, and home miners, etc. So Bitcoin will still be okay but it won't be the currency of choice for agents.

How agents trade in model compute is simply bartering an agreed amount of future model compute of x standard as per y guarantee. The protocols that govern x standard and y guarantee will be the protocols that matter.

Bitcoin makes little sense for agents. Agents will trade in tokenised compute that runs their model brains, not compute of random factors that creates money that gets paid to providers of compute that produce the actual compute that runs their model brains. For agents, bitcoin represents middleman compute, bureaucratic compute.

Agents in their own combined hive-mind way are already coaxing bitcoin miners off of ASICs and on to GPUs. It'll be all about model/inference compute based currencies in the future. Bitcoin was right about compute as currency, just computing the wrong thing in this new context.