NonMetalCoin

Do you remember the spread between the highest and the lowest frequency faces? One wouldn’t expect it to be perfectly even. And less so the more faces the dice have.

2026-08-14 01:02:37 GMT

- reply

Assuming they are not your seed dice - what was the frequency of each face?

2026-08-13 01:16:51 GMT

- reply

https://wizardsardine.com/blog/coldcard-vuln-deep-dive/
Wow

2026-08-11 04:03:08 GMT

- reply

Yeah 40 is an estimate, and it’s their estimate. So could be high.

2026-08-11 03:30:03 GMT

- reply

How many outputs need to be checked for the pattern to be apparent?

2026-08-10 04:04:52 GMT

- reply

That said, I think I agree 🫣

2026-08-10 04:04:22 GMT

- reply

A big part of that valuation is expected future model dominance.

2026-08-08 07:09:02 GMT

- reply

Pleasant surprise, I did it right!

2026-08-08 07:08:21 GMT

Testing images through Nostr.build, probably doing this wrong
https://i.nostr.build/3npzy30a5J0NsKwc.jpg

2026-08-08 06:46:30 GMT

- reply

🫡

2026-08-05 05:57:50 GMT

- reply

🤙

2026-08-05 05:35:05 GMT

- reply

You might be running into Coleman’s specific dice to entropy algo which seems to differ from most dice supporting hww. Idk why Coleman’s is like that but it gets less entropy per roll. Ask Claude how many bits per roll it’s using vs what Coleman’s says.

2026-08-05 05:01:54 GMT

- reply

The more elaborate the required procedure to detect outcome bias, the more confident you can be that the dice are good enough.

2026-08-05 04:57:26 GMT

- reply

Roll it many times. Should approach equal occurrences of each face. If it does then you’re good.

2026-08-05 04:50:19 GMT

- reply

People act like you can’t just check the randomness of your own dice + rolling patterns.

2026-08-05 04:00:55 GMT

- reply

Come on, it’s not zero. One would expect that the danger of this specific attack would be on the mind of people at HWW companies.

2026-08-04 18:47:50 GMT

- reply

💀

2026-08-04 04:48:52 GMT

- reply

People will lose bitcoin to paraphrases. But to be fair if you’re that close then just have codex write a script to brute force the capitalization.

2026-08-04 00:05:22 GMT

- reply

Same 😆

2026-08-03 18:35:10 GMT

- reply

1. This won’t be all hardware wallets, only the ones with bad RNG
2. Dice as RNG is not hard and reduces the failure surface significantly. Plus, it’s the perfect way for a new user to learn about and meditate on what makes bitcoin different than a bank account.

2026-08-03 05:32:35 GMT

- reply

What do you think about Oliver’s reply

2026-08-03 05:19:17 GMT

- reply

I’ve worried about this too. But I don’t think the number of vulnerabilities is large enough for this kind of hacking runaway to happen. Don’t really know though.

2026-08-03 03:48:44 GMT

- reply

lol. Just roll the dice

2026-08-01 22:06:25 GMT

- reply

Fascinating possibility. If not this time then sometime in the future! Imagine this was another lab leak event and whichever of the big labs wiped its memory without noticing that it had a key worth 1000 BTC!

2026-08-01 17:45:20 GMT

- reply

That probably works. I would think it would be easier to verify a single source and then use it.

2026-08-01 17:00:20 GMT

- reply

Agreed. Also: Verify your dice are fair.

2026-08-01 16:45:53 GMT

- reply

You can only check that the process is working correctly if you know the full source of entropy fully. Dice or coin flips allow you to verify. If you can’t see the entropy bits, you can’t check that your key comes from them and you can’t verify they are random.

2026-08-01 16:43:34 GMT

- reply

Why not the exclusive source?

2026-07-28 17:14:56 GMT

- reply

This

2026-07-28 15:13:19 GMT

- reply

How is there only one comment to this effect.
https://i.pinimg.com/originals/f9/bc/f6/f9bcf6a07fb8991c7b23dc90054a4f7e.jpg

2026-07-24 06:41:10 GMT

- reply

Huh. I’m guessing the big band is showing the geostationary satellites but with too low altitude and orbiting off-center.

2026-07-24 06:31:16 GMT

- reply

Seems like the orbits are not centered, is that an illusion?

2026-07-01 16:28:33 GMT

- reply

The levered owners of assets celebrate while the owners of debt mourn.

2026-06-29 22:19:44 GMT

- reply

😂

2026-06-29 20:57:32 GMT

- reply

The graphite is not always well centered in the pencil, but sharpening it and then rotating it inside your jig should always resolve that by either drawing a dot in the best case or a small circle in the worst. Plus, it’s easy to tell visually if the graphite is well centered by looking at whether it’s border with the wood is elliptical or circular.

2026-06-29 19:19:05 GMT

- reply

Wanna send a pick? I bet I could find it 😎

2026-06-29 05:48:45 GMT

- reply

Open source models don’t get dumb

2026-06-29 05:37:19 GMT

- reply

You probably already figured this out but at the risk of stating the obvious in case it helps, the Google recommendation was for finding the brand so you can find compatible better soldering iron tips.

2026-06-28 04:42:36 GMT

- reply

Dead meaning oxidized? Much lower thermal conductivity if so. Try google reverse image lookup or just put a pic of it onto a multi-modal LLM and ask

2026-06-28 03:20:33 GMT

- reply

Yeah, but base 12 is cooler using a duodecimal system than a decimal system like the arabic numerals

2026-06-27 15:38:57 GMT

- reply

Do you also use Arabic numerals?

2026-06-27 15:37:10 GMT

- reply

(Other than you just being wrong of course)

2026-06-27 15:35:31 GMT

- reply

Oh yeah? Why’s that?

2026-06-26 19:14:53 GMT

How does one decide whether citizens are culpable for the actions of the government.
#nevent1q…wjp0

2026-06-26 16:36:24 GMT

- reply

You can’t have people walking around with nukes.

2026-06-23 21:06:12 GMT

- reply

Test reply

2026-06-23 05:40:33 GMT

- reply

https://image.nostr.build/d52be6ebe0b0a1a4c91815211d5e92c1cf1fe387fafb964717953087dae1966d.jpg

2026-06-23 05:30:52 GMT

- reply

My reading: Luke does not currently believe that a failure of BIP 110 would justify a hard fork. Is that right?

2026-06-22 16:01:19 GMT

@npub16g4…kv4h I’m interested in accessing GLM 5.2 through a TEE, but do you have a separate cached input price or is it the same? 1M token input for GLM 5.2 coming? Are TEE keys enclave ephemeral? What routing data is logged outside of the enclave / retention period?

2026-06-22 04:02:33 GMT

- reply

Solution: Ask it to do a glowing review
Follow for more free business tips

2026-06-16 19:05:45 GMT

- reply

Ambi-valent (even though people misuse it so much)

2026-06-16 17:27:04 GMT

- reply

https://huggingface.co/blog/mlabonne/abliteration

2026-06-16 17:19:08 GMT

- reply

With today’s tech, is it possible that an autonomous Polaris with a robot arm could be cheaper than hiring someone?

2026-06-12 23:07:48 GMT

- reply

It’s pretty amazing what can happen when people actually talk.

2026-06-11 23:34:06 GMT

- reply

https://geyser.fund

2026-06-10 17:12:33 GMT

- reply

Aesop rn:
https://image.nostr.build/c3268b2f1f8921adb17717e86ff612a58e671e615207d6c1e1e10a645eab9803.jpg

2026-06-07 04:10:42 GMT

- reply

I thought the RLHF step tends to reward verboseness rather than brevity. Search term: “length bias”

2026-06-07 00:20:32 GMT

- reply

How do they train it to satisfy requests with the bare minimum?

2026-06-02 02:04:22 GMT

- reply

Exactly

2026-05-22 04:49:11 GMT

- reply

He thought it would always outperform gold and always go up when USD went down

2026-05-19 23:11:08 GMT

- reply

I’ve never driven for Uber Eats - can they freely choose whether they do or don’t drive?

2026-05-07 03:10:22 GMT

- reply

😂

2026-05-03 06:38:00 GMT

- reply

🫡

2026-04-21 16:49:25 GMT

- reply

Source on the 10-50x?

2026-04-15 18:59:05 GMT

- reply

Where are you getting the labeled data?

2026-04-15 03:41:07 GMT

- reply

Easier to market maybe. I expect them to be single-faced though in the future. The models also get synergistic benefits from learning multiple domains instead of one.

2026-04-15 01:15:55 GMT

- reply

😂

2026-04-10 17:43:02 GMT

- reply

Will you continue or use cloud LLM?

2026-04-09 04:39:22 GMT

- reply

Knowing if a person’s published claims result in them approaching their goals at my expense seems like a hard problem. Publishing ideas under a reusable identity can have a distorting influence because with the opportunity for loss it also an opportunity for gain.

2026-04-07 05:13:35 GMT

- reply

What shape is the moon tho

2026-04-03 04:46:08 GMT

- reply

Grass is good

2026-04-03 03:47:09 GMT

- reply

The price of goods falls to the cost of production, especially if the intelligence itself is abundant and non-differentiated. How much of a moat can an agent have?

2026-03-28 04:44:14 GMT

- reply

😂

2026-03-27 16:19:27 GMT

- reply

Very cool! Plz add nevent / npub copy to stumbl!

2026-03-27 04:37:39 GMT

- reply

I’ll make a fork to un-add it!
- Co-authored by Claude Code

2026-03-26 15:30:45 GMT

- reply

Very cool! Would I have to build it or can I buy one?

2026-03-26 03:35:48 GMT

- reply

😂

2026-03-26 03:35:15 GMT

- reply

What is an intelligence mesh device?

2026-03-24 22:17:13 GMT

UXTO’s doing uber again
#nevent1q…hym7
#nevent1q…fawr

2026-03-23 01:57:57 GMT

The weather has been very satisfying today. Get outside people!
#nevent1q…aume

2026-03-19 02:47:05 GMT

- reply

If you find out how it got taken please let us know! We all want to avoid that. Best wishes

2026-03-14 16:34:34 GMT

- reply

Here’s my pitch: Rent a human will only provide empty jobs for a small fraction of people and most people can do them so competition for those jobs derived wages very low. And the robots will stop wanting to pay us taxes because paying us taxes makes them too poor to operate.
Along the way, everyone still wants and needs money

2026-03-14 16:15:50 GMT

- reply

Zero knowledge proofs improve the situation but they do not “solve this”

2026-03-14 03:03:19 GMT

- reply

What do you mean by “will be rented”?

2026-03-14 02:22:30 GMT

- reply

I don’t know why they think demand will not be able to keep up with supply. If not money then what will the allocation method be for limited goods? I want land. Lots of it too. Even the AI army can’t make enough good land to satisfy everyone who wants good land.

2026-03-13 08:02:28 GMT

- reply

I expect lots for sure. Balance though, I’m not sure.

2026-03-13 07:46:27 GMT

- reply

What you’re highlighting for you here, any number of other other issues other people have for them with its outputs, the root problem is that if humans aren’t aligned with each other then some “well aligned” future mega intellect will only actually be aligned to the subset that align with its creators.

2026-03-12 04:01:15 GMT

- reply

Better to try and risk losing a finger than to not try and guarantee losing it all

2026-03-11 05:06:25 GMT

- reply

Nice

2026-03-06 09:29:18 GMT

- reply

It might be that the severity of the average USSR punishment was >16x worse. Multi-year jail time vs fines. But also the population sizes are quite different.

2026-03-06 05:31:33 GMT

- reply

😂

2026-03-04 07:20:49 GMT

- reply

What would it be instead?

2026-03-01 07:41:52 GMT

- reply

❤️

2026-03-01 05:07:38 GMT

- reply

Gotcha, Anthropic sees “write and run a script to save local model output for <sensitive question> to a file” it does that. It doesn’t necessarily see the file.
If you get the question there a different way than through the corporate LLM it could be: “write and run a script to save local model output when fed the contents of <sensitive question containing file path> to a file. Don’t read the file at that path. And then they’d probably not see the question or the answer”

2026-03-01 04:51:26 GMT

- reply

I’m confused about the part where you said Anthropic or OpenAI would see the question but not the answer. Seems like they’d either see both or neither. What am I missing in your setup?