semisol

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

2026-08-08 19:35:37 UTC

- reply

Secure element RNGs protect billions of credit cards, IDs, passports, computers, and more. Similar RNGs on HSMs protect the entire CA system, DNSSEC, and a lot of cryptocurrency. The Linux, macOS and Windows RNGs protect every website and application on the internet.
With secure elements, the only catastrophe that I can easily recall is RoCA. And that was caused by using a insecure prime generation algorithm instead of an RNG failure.
This is a uniquely Bitcoin problem. All the good infosec people work for general cryptocurrency companies, or at way more reputable companies.
With cutting corners and the incompetence of people at these companies, it is not unexpected that we have scared all the competent people away

2026-08-08 19:15:12 UTC

Regardless of your stance on spam, BIP-110 “UA”SF was a horrible idea. It will likely fail anyway.
The only thing it achieved is confuse people and polarize them further. Congratulations.
(My opinion is that we should not encourage it, by blocking it via *relay policy*. People don’t want to build on uncertain ground. But consensus should not be changed.)

2026-08-08 19:12:57 UTC

- reply

You sure it was not the so-so scary john-doe/qwen3.6-27b-ultra-heretic-abliterated-vl-nvfp4-q4-fable-distill-lora-merge-v2-gguf ?

2026-08-08 19:09:26 UTC

is anyone selling popcorn for sats?
#nevent1q…y0kj

2026-08-08 19:06:40 UTC

- reply

This is OpenAI/Anthropic

2026-08-08 05:42:58 UTC

- reply

Sorry I don’t think that’s gonna cut it. Buy a hardware wallet or two with the rest of your sats, so you can make a 2-of-3 to protect the 0.2 sats you have at the end

2026-08-08 05:35:05 UTC

Rule 1 of security research: until it is fixed and properly disclosed, you NEVER:
- share the details
- post what you found a vuln in
- actually, just keep your mouth shut and don’t say a word

2026-08-08 05:01:58 UTC

1. Make new model
2. Release
3. Quantize and lobotomize it
4. goto 1

2026-08-08 04:09:34 UTC

Am I even surprised
#nevent1q…xgrv

2026-08-08 04:07:09 UTC

- reply

and codec, serdes, …

2026-08-07 20:51:49 UTC

- reply

The certificate is for the chip. It does not mean whether the chip is used properly by Ledger, Trezor, etc.
It only says “the chip CAN BE secure”, but it is very easy to undermine the chip’s security as the developer.

2026-08-07 19:19:38 UTC

- reply

No, because the certification means absolutely nothing for the HWW’s security. It’s for the chip only

2026-08-07 19:14:08 UTC

Note that EAL5/6 certification means absolutely nothing about the security of a hardware wallet,

2026-08-07 19:00:58 UTC

- reply

that changes absolutely nothing

2026-08-07 11:10:24 UTC

- reply

the translation is accurate
https://i.nostr.build/7tHsaJSyVUZUkTCN.jpg

2026-08-07 09:34:33 UTC

- reply

actually this assumes the person fabricating it can even use photoshop, no, it’s a dumb copy and paste in paint

2026-08-07 09:33:59 UTC

- reply

it cannot be, and that image is clearly photoshopped

2026-08-07 09:33:34 UTC

- reply

It does not take time to do CVSS scoring, and the CVSS scoring system is known to have issues many issues anyway.
1. Scores are often much lower or higher than they should be.
2. It is possible to easily over- or understate the impact of a vulnerability, intentionally or not.
3. Whether the vulnerability is actively exploitable is another question.
After NIST had slowed down enrichment of CVEs, and many other safeguards broke, anyone can now go and issue a CVE for a project without any verification.
https://sqlite.org/forum/forumpost/34bdf3b9bd759d4d

2026-08-07 08:14:27 UTC

It seems that many people on Nostr just want to pretend to be devs just for grant money and nothing else.

2026-08-07 07:13:19 UTC

In cryptography the best way to check if you are sufficiently skilled yet, is if you feel like you know a lot but also nothing at the same time

2026-08-07 07:11:19 UTC

- reply

If you want to play both chains, you’ll need to use a UTXO that is an descendant of a new block’s coinbase

2026-08-07 07:10:23 UTC

- reply

To not care

2026-08-07 07:08:01 UTC

- reply

And what then? They replayed your TXs on BIP110, congrats, nothing changes on the main chain for you

2026-08-07 06:33:18 UTC

- reply

Well it a fork could make you lose LN sats, so could a malicious actor

2026-08-07 06:00:05 UTC

gm Nostr
If you are concerned about BIP-110: don’t be
You cannot lose sats unless you try to play both sides of the fork. It is impossible.

2026-08-06 23:16:35 UTC

- reply

Anything can be a critical CVE if you exaggerate it enough

2026-08-06 19:04:31 UTC

- reply

Well a harness that wastes tokens is a great way to maximize money spent

2026-08-06 18:14:39 UTC

About this entire AI security scanning thing:
1. Why are people using $ spent as a metric?
2. A huge chunk of these vulnerabilities are likely hallucinated and/or overstated
3. You can’t just point AI at a pile of files and expect good results

2026-08-06 00:05:42 UTC

- reply

wtf

2026-08-05 21:42:16 UTC

- reply

This happens

2026-08-05 21:28:21 UTC

Warrant canaries are security theater

2026-08-05 14:46:14 UTC

- reply

Card shuffles are not very random processes.

2026-08-05 13:52:54 UTC

I find my feed is better if I immediately mute anyone who posts AI slop.
GM.

2026-08-05 12:54:37 UTC

- reply

This is very easy to get wrong
Do not recommend, do not do this, avoid, just run, please