<unnamed>

Public Key

npub12262qa4uhw7u8gdwlgmntqtv7aye8vdcmvszkqwgs0zchel6mz7s6cgrkj

Profile Code

nprofile1qqs99d9qw67th0wr5xh05de4s9k0wjvnkxudkgptq8yg83vtulad30gpp4mhxue69uhkummn9ekx7mqppamhxue69uhkummnw3ezumt0d5r7zqwm

Publishing to

nostr-pub.wellorder.net

Last Notes

2026-09-16 00:04:28 UTC

- reply

This is also a factor in Nostr.land. The utilization is not very high, so if I pushed that more, it would be eveb more efficient.

2026-09-15 22:26:41 UTC

- reply

Well, Nostr.land is a much larger operation for example, with aggregation and such taking a lot more resources. Also one of the more popular relays.
Doing a relative comparison, $50/mo is grossly inefficient for that size.

2026-09-15 01:33:15 UTC

- reply

I would not run one

2026-09-14 20:09:13 UTC

- reply

Still have more bits to quantize

2026-09-14 19:52:11 UTC

My node is down starting immediately to switch to Eclair so I can sleep well at night

2026-09-14 01:52:50 UTC

- reply

pls pipe to /dev/null🤣
if we will have LLM generated “memes” they should be required to be generated with a model no more capable than GPT-2

2026-09-14 01:07:19 UTC

The only people that I see excited about AI generated images, video, apps, and slop-posts is people too deep into the AI delusion.
Denial?
Instead of a slop video or image, you could have just written out the prompt you used. Much better.

2026-09-13 23:36:28 UTC

I have received more in Monero, in terms of both transaction count and total value, compared to zaps.
Even excluding outliers, it is similar.

2026-09-13 23:35:33 UTC

- reply

We have finally created the Torment Nexus from the book Don’t Create The Torment Nexus

2026-09-13 17:37:42 UTC

- reply

Any proof of this? I still get a normal URL with query params and the video ID

2026-09-13 13:08:43 UTC

- reply

I feel this most in Bitcoiners

2026-09-12 18:23:08 UTC

- reply

It is not

2026-09-12 16:33:32 UTC

- reply

The hidden states across layers are huge, especially for big models, for example 32KB per token. Especially for prefill, that means 3.2G of data per layer to do a prefill on 100K context

2026-09-12 15:44:45 UTC

- reply

Due to the architecture of LLMs it is practically impossible to have distributed compute beyond the same datacenter

2026-09-12 14:49:11 UTC

- reply

Yes.

2026-09-12 12:26:08 UTC

- reply

Standalone 🍍

2026-09-11 23:10:43 UTC

🍍?

2026-09-11 23:00:04 UTC

- reply

Yeah.
And it is also not a guarantee that Bitcoin (or any other asset) will succeed, or succeed for *you* at least.
Your goals may not be aligned with the majority of users, or there may be some catastrophe, or a better alternative.
It is better to diversify, but to do that with proper research.
Premined coins, random “tokens” and solutions that bring nothing new to the table don’t make sense.

2026-09-11 22:46:38 UTC

- reply

It means the NWC connection was lost and is falling back to the native wallet

2026-09-11 22:46:06 UTC

A specific currency is only a means to an end, and you should use what is best for your usecsse
There is no one solution to rule them all
(This can also generalize to anything.)
#nevent1q…t7r5

2026-09-11 19:49:18 UTC

- reply

All LN software hates their users

2026-09-11 14:57:01 UTC

I have had a Monero payment target added for a while.

2026-09-11 02:17:57 UTC

- reply

Yeah. But you can still be sent a fake email from a genuine address if their outbound server is hacked.
I would expect that any competent team uses something like AWS SES rather than some cheapo provider. Or they self-host.
Also, require DKIM, *and keep your own keys*.

2026-09-10 18:23:26 UTC

- reply

And this is why if you outsource your mail you do it properly.

2026-09-09 15:16:20 UTC

- reply

Well, drives do not need as much cooling and can be densely packed

2026-09-09 14:51:05 UTC

- reply

The data centers for those are much smaller, I think.

2026-09-09 08:45:17 UTC

- reply

From my knowledge there is not much to cache? The range-proof would be TX specific anywayz
The only caee it might benefit is mempool / block validation; which is so dumb for a network that could proceed all txs on a toaster

2026-09-09 00:45:06 UTC

- reply

About the 1-of-5 let’s be honest about this, you’re supposed to do multiple rounds to turn it into 1-of-practically all CJs.
But default privacy still wins + the anonset. wen FCMP++?

2026-09-08 19:43:50 UTC

- reply

Why are one-time-use proofs being cached? Why is the cache key exactly not what is being proven ??

2026-09-08 17:38:00 UTC

- reply

I think it may be Proton that is auto-highlighting the URL “Claude.ai” instead

2026-09-07 17:09:13 UTC

- reply

already done. I’m trying to build a corpus of clanker made and non-clanker made apps to see if I can make an AI detector for code

2026-09-07 15:56:51 UTC

- reply

most of the code should be clanker generated

2026-09-07 15:18:48 UTC

Good morning Nostr
I am looking for clanker generated Nostr apps

2026-09-04 22:32:27 UTC

GPT 6 Astra is misaligned
especially with the user
It refuses to touch “private” “internal” CoT of local models where I can… see the CoT?

2026-09-03 18:09:28 UTC

“gated” is an easy way to identify people trying to cosplay as a SWE

2026-09-02 22:13:45 UTC

- reply

Oh, inbox.

2026-09-02 20:46:52 UTC

- reply

It would be stupid to remove hist.

2026-09-02 20:02:23 UTC

- reply

Would recommend trying nostr.land. If you have any issues let me know

2026-08-31 14:14:08 UTC

- reply

worse

2026-08-31 14:13:54 UTC

- reply

I’d hope the AI misalignment researchers are on this one, since the model is basically misaligned with the interests of everyone (excl. Anthropic) including the user
Oh wait they are trying to intentionally train this into the model?!

2026-08-31 13:53:48 UTC

- reply

what the fuck is a “load bearing” and this is not even a primitive

2026-08-31 02:21:04 UTC

What is your biggest problem with Nostr relays or clients?

2026-08-31 02:13:21 UTC

- reply

LLMs allow a smaller quantity of labor to produce a greater *volume* of work

2026-08-29 00:00:26 UTC

🤔

2026-08-27 09:29:04 UTC

- reply

LND, and Lightning Labs’ behavior is actively harmful to the development of the Lightning Network.
They have refused to implement specifications such as Bolt12 and dual-funded channels.
It is also by far the most annoying node implementation I used. Not to mention that all the issues I had (including losing half of my capacity at one point) were with LND peers.

2026-08-27 07:38:12 UTC

Nostr.land sales will be stopped for now due to the CLN bug. In the longer term, I will probably move to Eclair.
If you would like to purchase during now, please reply to this post (no DMs)

2026-08-25 22:49:16 UTC

- reply

It shares many methods with the attempted backdooring of xz

2026-08-25 19:06:49 UTC

- reply

why unedited clanker slop ⁉️

2026-08-25 16:06:43 UTC

Good morning Nostr
I believe that relays and media hosts should be opinionated, but should also be fair. Most importantly:
- Be honest about your policy (even “whatever I don’t like” is better than most)
- Apply them as you stated
- Do not try to sabotage your users silently
- Except in egregious circumstances, help them move out

2026-08-25 13:55:15 UTC

- reply

all the commercial NFC chips on the market are garbage in terms of security

2026-08-25 12:57:00 UTC

- reply

what specifically

2026-08-23 18:44:31 UTC

2026-08-23 02:48:46 UTC

Note: very long
https://elan.school

2026-08-22 22:21:14 UTC

- reply

We have at least the illusion of consensus on SegWit, and that a good chunk (not any majority) of economic participants do enforce it (considering they rely on for example witness-embedding TX submissions)
I guess this all inevitably reflects in miner actions, through nth-order effects.

2026-08-22 19:17:57 UTC

- reply

The reason a UASF can work is that the general economic weight and public balance can shift significantly against the miners, which cannot tolerate the revenue loss (immediate, and longer-term reputation damage to “Bitcoin” itself) that would arise from even a few weeks of conflict.
Usually by solving an important problem and being unresolvable via any other way.
BIP-110 was not technically sound, did not manage to convince many economic participants, and had no benefit to a majority of people. It was also rushed, giving little time for a possible UASF to be even reasonably possible. If anything, it was against the interests of many parties, even the “good” ones.
(Everyone is also strategizing about what everyone else will do, and so even just the appearance of broad consensus is critical to have a successful UASF)

2026-08-22 19:12:29 UTC

- reply

Bitcoin is not the heaviest valid chain, and there is no “definition” of Bitcoin either.
It is a vague term that exists because of a general consensus all sorts of participants, which have weight in different ways (economic, computational, leadership, publicity, etc.)
And there is no strict weight to define 50%+1 but many different powers that form a result.
A node by itself does not have any weight in this, and only its operator could hold weight by being economically active (buyer/seller/platform), mining, via public participation in discussions, etc.

2026-08-22 12:43:06 UTC

You cannot expect “creativity” from an LLM. It can only create “new” things via random search.
Any “gaps” that need to be filled from input => output will be filled with a somewhat randomized average of what already exists. It will not add any new substance.
#nevent1q…pnn8

2026-08-22 12:36:46 UTC

- reply

It is 99.99% air and 0.01% (mis)info

2026-08-22 09:34:55 UTC

AI generated “content” (slop) is a shell of the real thing. It *looks* sensible but it’s empty inside. It’s worthless and doesn’t make sense.

2026-08-21 07:20:14 UTC

2026-08-21 04:39:34 UTC

- reply

Why build the Torment Nexus from the book Don’t Build the Torment Nexus?

2026-08-20 07:12:49 UTC

javascript was a mistake

2026-08-19 14:55:30 UTC

- reply

it is (properly) called a convection oven

2026-08-17 17:52:39 UTC

The AI psychosis gets stronger by the minute..

2026-08-17 12:19:04 UTC

- reply

The latter is also not true.

2026-08-16 23:08:53 UTC

- reply

Y’all are visiting websites?

2026-08-13 20:04:08 UTC

- reply

Yeah, do it

2026-08-12 23:47:27 UTC

- reply

I have a hard time taking Passport Prime seriously after several of their claims and decisions. The idea makes sense, but at that point Betrusted seems like a better option

2026-08-12 08:49:05 UTC

expand 32-byte k

2026-08-12 03:53:27 UTC

- reply

I agree that macOS is pretty great in terms of security model, but that desktop is in general a mess …
I’m more curious about your opinion for more limited tasks, and not day-to-day usage, like hardened limited-purpose servers, or to process sensitive data, and such.
It is unfortunate we don’t have a lot of investment in this area.

2026-08-12 01:15:13 UTC

- reply

I am aware :) I need something that works on X86 systems,

2026-08-12 01:14:44 UTC

- reply

Ideally, we have a microkernel, and there is no dom0. The problem is that you are now reinventing the universe.

2026-08-11 23:01:33 UTC

- reply

I think that dom0 should stay a vegetable

2026-08-11 20:51:32 UTC

- reply

And I guess, it is more about the security configuration than anything, and distros. Very few things have SELinux properly configured, for example

2026-08-11 20:50:55 UTC

- reply

What would you say is a good desktop OS?

2026-08-11 19:13:58 UTC

- reply

Yeah. This is a common method, used by basically anyone serious (and not).

2026-08-11 18:29:21 UTC

- reply

This is what it is for. So that people do not bruteforce the candidates ahead of time to manipulate it

2026-08-11 17:21:18 UTC

- reply

People comparing it to UASF and saying “bitcoin dead wallst took over” need to understand this. I think that no one would support a rushed attempt with threats of forking off *without any plan*, especially when the solution did not fix the problem.
Can we just have relay policy already?!

2026-08-11 15:53:29 UTC

- reply

IIRC, RandomX has been stable for a long while.

2026-08-11 12:54:22 UTC

- reply

After post-training, the model’s output distribution collapses a ton.
I feel like the recent models though are extremely annoying to use. GPT 3/3.5/4/4o/4.1/4.5 and even 5 was fine. There were patterns (like everything being a list for 3-series) but nothing this annoying.
Ever since they started optimizing thinking and especially coding, these models have gone down the shitter in terms of their “writing style”

2026-08-11 12:49:47 UTC

- reply

RLHF

2026-08-11 00:54:58 UTC

- reply

it is literally claude 5 speak
either it is ai written or someone took over @npub137c…k37w's account

2026-08-10 22:00:22 UTC

FCMP++ is all you need

2026-08-10 13:26:28 UTC

- reply

It was a post by @npub1f5k…vk2u about https://en.wikipedia.org/wiki/Nobel_disease

2026-08-10 13:21:59 UTC

This is what happened to all the Nostr influencers I think
#nevent1q…t24p

2026-08-10 11:35:30 UTC

- reply

Already solved: Arweave

2026-08-10 01:42:52 UTC

- reply

pci compliance headaches in 3.. 2.. 1..

2026-08-09 22:35:33 UTC

- reply

Breaking news: batching more efficient than not batching
(I’d recommend more aggressively batching, and also looking at using speculative decoding if you don’t already)

2026-08-09 02:03:52 UTC

- reply

The more complicated, the more likely you are to lose it

2026-08-09 01:56:37 UTC

- reply

none of those, you should be doing [under NDA] with [under NDA]

2026-08-09 01:54:07 UTC

- reply

[under NDA]

2026-08-09 01:52:17 UTC

- reply

it is not bad, but to get proper functionality on smartcards, you need specialized cards (oversimplifying a whole lot)
and getting those is very very hard

2026-08-09 01:48:55 UTC

- reply

not secure at all
what kind of nfc card

2026-08-09 01:47:38 UTC

Seedless hardware wallets 🤔

2026-08-09 01:42:10 UTC

- reply

I can’t even see the conversation beyond @npub1m50…glzq quoting “X coin go up”

2026-08-09 01:07:25 UTC

- reply

I have personally seen this in hospitals, and I think even some hotels

2026-08-08 20:45:44 UTC

BIP-110 has activated. What now? (not a lot, actually)
However, there is a very specific type of scam that could lead to you losing coins. Please read.
#naddr1qv…0kna

2026-08-08 19:48:41 UTC

- reply

Do you want to sell fork coins? (They will likely have no value anyway so not worth bothering)
Yes: you need to care
No: absolutely nothing changes except maybe some BIP110 peers need to have their channels force closed
Risk of fund loss: none

2026-08-08 19:39:24 UTC

and just like that, absolutely no one noticed
https://i.nostr.build/0KjAdVx2oLBxcU3I.jpg

2026-08-08 19:38:41 UTC

it is forking time