m0wer

JoinMarket NG

Public Key

npub1w3vaxva0vcrx7pnvlpmede5smvafdl69xnu7ma82kaxl9us89zdsht4c5c

NIP-05 Address

Profile Code

nprofile1qqs8gkwnxwhkvpn0qek0sauku6gdkw5klaznf70d7n4twn0j7grj3xcpp4mhxue69uhkummn9ekx7mqpz3mhxue69uhhyetvv9ujuerpd46hxtnfdujt4l6a

Publishing to

nos.lol

relay.damus.io

henhouse.social/relay

nostr.bitcoiner.social

nostr.sgn.space

relay.primal.net

Author Public Key

npub1w3vaxva0vcrx7pnvlpmede5smvafdl69xnu7ma82kaxl9us89zdsht4c5c

Show more details

Last Notes

2026-07-15 20:02:06 UTC

Decentralized lnproxy (provider discovery through Nostr) working on signet :-)
PR: https://github.com/lnproxy/lnproxy-relay/pull/31
Try it at: https://m0wer.github.io/lnproxy-webui2/?network=signet
BTW this is a nice web LN wallet for testing on signet: https://ln.signet.coinbin.org/

2026-07-15 16:38:29 UTC

Let me present you what could be the future of JoinMarket and hopefully a great tool for Bitcoin privacy and fungibility.
First of all, there is the motivation. JM makers have an onchain fingerprint, that although it's hard to follow, limits the privacy achieved by the takers. The full details are at https://gist.github.com/m0wer/a228c625fcb6a27c32e298ec903dfc44
But to save you some time, here the main problems described there are: unique enough maker fees, co-spending of inputs, makers never spending, ... Which in the worst case scenario, would mean that a pure onchain observer could potentially cluster some makers across mixdepths and reduce the anonymity set by one of CoinJoins where they participate.
So what can we do? Well, we can use this chance to bring JM to the next level. And create such a transaction graph "mess" that becomes a chain analyst nightmare :-)
Let me explain. We can't ever get fully rid of the links between UTXOs onchain, or at least not without a way more complex multi taker model that leads to Knapsack "dense" transactions. But we can strive for something, that is that no single party ever has all the information.
The idea is the following. JM makers could become Lightning Network swap providers. Like Boltz, like SwapMarket, like electrum swapserver. Using Taproot swaps and advertising through Nostr. LN swap users get a simple and trustless client side web UI, makers get "fresh" UTXOs from onchain->LN swaps and the opportunity to "spend" CJ outputs for LN->onchain swaps. Oh, and another source of fees. LN swap users get cheaper swaps from the increased competition.
But this is not enough. We need to break the subset sum analysis onchain. We want to stop maker clustering from the root. Here is the next idea, JM CJ participants (makers and taker) could get their change as an LN channel open. Or two, with random other peers and the change amount split randomly between both channels. So a maker that has a 10 M sats input in a 5 M sats CJ no longer gets just a clear ~5M sats change, instead, two LN channel opens where the balance is hidden. Maybe 2M sats in one and 3M sats in the other. But an external observer wouldn't know. Not even which channel is whose!
You could say, well, but an active attacker could probe those channels balance. Maybe, but not really. Those channels would be private simple taproot (HTLC based, not PTLC yet unfortunately), so not advertised to the network. And, BTW, onchain they look like any other P2TR output BTW. Advertised only when needed with a SCID alias that does not reveal the onchain UTXO. But hey, who knows, maybe some peer node that gets the SCID could try to probe it. Fear not! There's another ingredient coming to the mix.
lnproxy is a great service to hide the destination node of an LN BOLT11 invoice by having a relay/proxy node wrap it. The payment hash is the same, only the recipient can settle the payment. But whoever sees the wrapped invoice sees the relay/proxy pubkey instead of the destination one.
Why not decentralize lnproxy through Nostr and then makers can be lnproxy providers as well? Just to add a little bit more noise to their LN activity and help move the balances around. So that if these change as channels ever close (cooperatively) the onchain fingerprint those not help understand what happened in the opening CJ. Or that UTXO can still be used as CJ inputs if signed cooperatively. Or, of course, as LN channels for swaps of lnproxy. Some kind of noisy spin of CoinjoinXT.
What we need for this and what we're currently working on at jm-ng:
- A JM taproot only pit.
- Decentralized lnproxy.
- Decentralized LN swap providers.
- Protocol wiring for the change as channel CJ coordination.
- People to actually use this and not be overwhelmed by the complexity.
But if we manage to make this work, we will have awesome onchain and LN privacy. And a way for service providers to earn money anonymously by providing privacy to others. So it's definitely worth a try :-)

2026-07-14 14:49:07 UTC

https://github.com/joinmarket-ng/joinmarket-ng/releases/tag/0.34.0
Automatic history import for existing wallets, fee quantization defaults for better privacy, coin control labels, local mempool API support, bug fixes, and security hardening.

2026-07-14 12:26:58 UTC

- reply

I would guess that's just a fee reserve right? Not the actual cost.

2026-07-12 14:38:40 UTC

- reply

So, what if JM makers "gambled" their change? Maybe chunks of it?
Just a random idea: Could the probabilities be adjusted so that if maker A that "has" 1M sats change and maker B that "has" 3M sats, instead get proportional chances (¼ and ¾ respectively) to win 4M sats? Maybe they're willing to play this game for a expected value of 0 but a positive expected value of privacy gain.
I guess it doesn't work with more than 2 participants, but if it did, I'm sure many JM takers would be willing to gamble their change, for either a 10x that makes it worth it to deal with it (toxic change), or just lose it.

2026-07-09 22:49:03 UTC

- reply

superspeeded by #nevent1q…ydlk

2026-07-09 19:57:34 UTC

Wasabi coordinator tagging attack explanation and PoC: https://gist.github.com/m0wer/2e2f0a3241d8e7651ffdda4435e729ec

2026-07-09 13:01:40 UTC

You are fully trusting the coordinator with your privacy when using Whirpool (Samourai). Here are some practical attacks with a PoC.
https://gist.github.com/m0wer/e701bf5934474935c459b5e84e5d4e1b

2026-07-07 17:46:48 UTC

- reply

In many cases the default variance is not enough to overlap with any other maker (and even when it is, the anonymity set might not be great).
When several maker do overlap in fees, randomization becomes "harmful" IMO. Because it then helps track an individual maker by following their "randomly unique" fee through the orderbook.
This second part is arguable. But for a simple message to reach makers and give them an "easy" fix, I think it is better to not get into much detail (but the links expand on this for anyone interested).

2026-07-07 11:57:46 UTC

JoinMarket makers with unique enough fees are easily clustereable across mixdepths and don't provide much privacy to takers that choose them.
Imagine there is only one maker that charges 0.123% fee. An external observer can identify that maker's equal amount outputs when they are provided as inputs for another coinjoin with the same fee (observable onchain).
So I would encourage makers to select one of these quantized fees and disable randomization. This is an opt in change that joinmarket-clientserver and joinmarket-ng makers can apply on their own.
The more makers on the same quantized fee, the better anonymity for them and consequently for takers that pick them.
More info at https://joinmarket-ng.github.io/coinjoin-simulator/mainnet-deanon.html and https://github.com/joinmarket-ng/joinmarket-ng/issues/508
https://blossom.primal.net/85c865af60dfba6d9e990b7a700fee698e37310c0c7f0fd6dcb724ea24ce1696.jpg

2026-07-05 14:41:19 UTC

JoinMarket makers can be clustered (across mixdepths). Two ways:
- Easily if their CJ fee is unique enough and reused between rounds.
- Harder by gathering co-spending thought many rounds and waiting for the intersection to be narrow enough.
It's easier said than done in practice. Bit it's still something to improve.
Here is a more detailed explanation of the problem and possible solutions that could bring JM achieved anonymity to the next level.
https://gist.github.com/m0wer/a228c625fcb6a27c32e298ec903dfc44
Feedback welcome!

2026-07-05 14:21:06 UTC

Nice podcast. Bitcoin is unstoppable, but you are not.
Privacy is what makes you not worth attacking.
https://youtu.be/0_hlkmpw1sk

2026-07-05 09:43:22 UTC

- reply

Tumbler (sweep) works in Jam V2 with jm-ng backed😎

2026-07-05 07:54:46 UTC

- reply

You said makers know "all the takers". I said makers know only as much as a passive onchain observer. Which is what that comment talks about.
The tumbler works, and its functionality can be achieved manually as well. Additionally, I've implemented options like https://github.com/joinmarket-ng/joinmarket-ng/pull/280 that hide the taker onchain fingerprint.
And currently working on many other improvements at the ecosystem level.
JM is not perfect and I'm happy to point out its flaws. Makers knowing "all the takers" is not one of them.

2026-07-04 09:54:48 UTC

- reply

It's important to note the JM is a market. There are offers with 0 fees. And offers with rather high fees. The question is how much Sybil attack risk you're willing to tolerate.
I honestly don't think Whirlpool is great... You really are trusting the coordinator with too many things: ensure fresh liquidity, not isolate all participants into Sybil rounds (it could just give a unique round ID to each request), and then there's also post mix cleanliness which I don't think most users do well. And with fixed denominations it's even harder to do that right.
Wasabi coordination is nicer IMO but there are similar tagging attacks the coordinator could do. And there's also the post mix problem.
Maybe I would also recommend it for newbies. Because risking privacy from the coordinator ia better than nothing.
But I really think the solution is a market and is decentralized. That's why I work in JM. It's almost impossible to prevent tagging attacks completely with central coordinators.
And BTW it looks the most used Wasabi coordinator runs behind a Cloudflare tunnel. Meaning Cloudflare could do some interesting connection timing analysis passively and link inputs together and even get some info on the output mapping. And of course it could selectively drop connections and so on. And since the round success rate is around 16% for large rounds (iirc). So that's really bad IMO.
Good thing of not making money from people using JM is that I'm free to talk about flaws and limitations. This is would be very hard to do for a company running a centralized coordinator whose existence depended on it being used.

2026-07-04 09:35:53 UTC

- reply

True.

2026-07-04 09:28:38 UTC

- reply

Yes! jm-ng is backwards compatible, so it can use the existing liquidity and provide it as well.
Here is an orderbook watcher that has more details about makers and directories: https://joinmarket-ng.sgn.space/

2026-07-04 09:24:43 UTC

- reply

In JM, makers don't know anything else about the taker other than what an external onchain observer sees, and their own input-output mapping (of course). The taker is the coordinator and signs last. Each maker sees the whole unsigned transaction and signs their inputs. JM has other "flaws" though.
In Wasabi and Samourai you are trusting the coordinator not to attack you. But it could. See: tagging attacks.
I think we would all be better off if Bitcoin privacy wasn't about "cults".

2026-07-01 11:29:45 UTC

- reply

Fair point. It was more of a clickbaity opinion than a rigorous idea.
With using it wrong I meant using AI to avoid thinking rather than using it to enhance thinking.
That's just one aspect of it, of course if there's a task that you don't enjoy doing and AI can do it good enough, it's logical to use it for that.

2026-07-01 10:51:51 UTC

- reply

I've never been more intellectually challenged in my life. AI has taken over all the "simple" tasks such as waiting for tests and fixing failures, dealing with dependencies, writing documentation, ... And I get to deal with the more complex stuff way more often than before.

2026-07-01 10:49:10 UTC

If AI is making you "think less", you are using it wrong.

2026-06-30 17:34:34 UTC

- reply

Sure. I think its goal is to explain the timeline and core concepts only.

2026-06-30 13:45:09 UTC

- reply

Blame @nprofile…kuag 🤪
I originally posted it there but then on nostr the markdown links are stripped out and appear at the bottom (at least on Primal).

2026-06-30 12:55:44 UTC

Jam V2 beta (JoinMarket WebUI)
https://github.com/joinmarket-webui/jam/releases/tag/v2.0.0-beta.0
TLDR: Big frontend rewrite with many improvements. Using [https://github.com/joinmarket-ng/joinmarket-ng/](https://github.com/joinmarket-ng/joinmarket-ng/) as the backend.
A standalone docker image for fearless beta testers is coming next.
Any and all feedback is highly appreciated.
Please know, that there are still open issues and refactorings needed before the final release.
https://stacker.news/items/1517235

2026-06-30 12:16:29 UTC

- reply

where was it?

2026-06-30 09:47:34 UTC

- reply

"""
~August 2026
Mandatory Lock-in
If not locked in early, mandatory signaling begins — blocks that don't signal are rejected as invalid, guaranteeing lock-in.
"""
bip110.org

2026-06-30 05:13:39 UTC

- reply

Sure, only if over the difficulty threshold. But that's what pools count anyway, and would be impractical to fake it. The same principle of mining itself.
But they have to be verified! The novelty here is that until now pools just cared about the coinbase transaction, now they might have to care about all! And I still don't know the answer to my original question of what will OCEAN do about that.

2026-06-30 05:09:25 UTC

- reply

Sure, but the ordinary "race" part is very relevant. Same will happen on the non BIP-110 side of things (that's what I meant).
In any case, a miner seeking profitability only, wants to be on a pool where all other miners share the same consensus rules. Otherwise your peers "wins" become your failures. And at the same time you want that pool to be on the side "winning" the race overall.
So I think there might be a shift of miners towards more centralized pools that enforce one side or the other.

2026-06-30 05:03:14 UTC

- reply

It's explained pretty well here: https://bip110.orange.surf/

2026-06-29 19:28:36 UTC

- reply

No idea, better ask tbk.

2026-06-29 19:26:52 UTC

- reply

Of course there is! What would be the point to mining otherwise!
The block hash includes the merkle root of the transactions, so it could be verified easily. And it must be checked! Otherwise how does the pool know to what address you are mining the reward to? Would a pool pay a miner that might be sending the block reward + fees to their own address? You don't need to disclose the whole merkle tree to prove that the coinbase transaction is correct. But the mechanism is there.
In the case of BIP-110 activation, I guess the BIP-110 miners wouldn't want to share rewards with miners that, if lucky, would include "invalid" transactions that they would later reject as a valod block. So then you need to check the whole merkle tree to recompute the root and validate the shares of all miners.

2026-06-29 19:15:17 UTC

- reply

Thanks for the answer. I meant this with the profitability question:
#nevent1q…jeve

2026-06-29 12:40:14 UTC

- reply

Oh and about profitability, the problem is not the small difference in fees depending on the block transactions. I was talking about having most of the pool's hashrate mining only over a BIP-110 chain of blocks (fork) that then get reorged so all those block rewards are gone and the hashrate "wasted". The impact of profitability would be huge in this case. And it's more about what people within the pool decide to do, not even conditional to overall support of BIP-110.

2026-06-29 12:36:10 UTC

- reply

DATUM allows for pooled mining while each miner decides on their block template. That's what I was referring to.
Currently, you can mine for Ocean pool and get paid the proportional share of your hashrate for the pool mined blocks while choosing your own block templates.
This gets interesting if some blocks stop being considered valid. Or if Ocean needs to check every miner's block template to decide if their shares are valid or not.
And Ocean is a company. That does way more than "organizing hashrate" so it's not crazy to ask them for their intentions on this topic.

2026-06-29 11:54:00 UTC

@nprofile…0rnh @nprofile…c93l Will OCEAN pool enforce BIP-110 compliance for DATUM miners? And for its stratum?
What happens to a DATUM miner that builds its own block templates if it finds a block that includes non compliant transactions? Or even if it doesn't find it, does the hashrate still count?
And of the pool enforces BIP-110 even if only for stratum, I guess rewards will come from BIP-110 blocks only. Which means lower profitability (but same would happen in the other side). But additionally means that many block rewarda are at risk of being invalidated (reorg). So what happens to LN instant payouts?
Don't these questions deserve an official announcement? Maybe I've missed it but couldn't find any.

2026-06-29 09:08:25 UTC

- reply

Frontend: big rewrite. Smoother and with new features, especially interesting for understanding failures.
Backend: v1 supports only joinmarket-clientserver (now archived) v2 supports JoinMarket NG which is maintained and provides new endpoints for Jam (history, logs, ...).

2026-06-29 05:49:47 UTC

Jam v2 (JoinMarket WebUI) beta is out! Adding compatibility for JoinMarket NG as the backend. See https://github.com/joinmarket-webui/jam/releases/tag/v2.0.0-beta.0
And @nprofile…79xu has audited and co-signed up to the latest release of JoinMarket NG: https://github.com/joinmarket-ng/joinmarket-ng/commit/947552bead929c342282019932ff12fe17cc44af
Huge milestones! Long live JoinMarket.

2026-06-25 21:11:28 UTC

- reply

https://blossom.primal.net/c1899ebb9449dcfdc537d556adb34e4c9244347b16df61d10e3d05eda1cff70f.jpg

2026-06-25 05:52:27 UTC

- reply

what is it?

2026-06-22 15:29:19 UTC

- reply

Would it be possible to create a bridge between Waze or similar and roadstr?

2026-06-21 07:32:43 UTC

- reply

Estamos pendientes de que salga la beta pronto, le queda muy poquito.

2026-06-21 07:32:10 UTC

- reply

sí porfa!

2026-06-21 07:31:16 UTC

- reply

I think the bigger problem there is needing a company to perform businesses. That's why "tainted" coins are a problem, not because they have anything wrong on their own. But even if you cared about that, LN fixes that. Blockchains don't scale anyway.

2026-06-21 07:26:42 UTC

- reply

Monero is still vulnerable to many attacks that collapse anonymity. That's why they're working on FCMP++. But even with that fixed, blockchains don't scale.

2026-06-21 07:10:46 UTC

- reply

Should I be angry at my grandma because of her public pension? She's receiving way more than she contributed.
Unfortunately many people benefit unfairly from the state. Should I also "hate" anyone that went to a public university or hospital?
I see the point, but don't think it's fair. I hate the state, which is clearly who steals from me. And I'm disappointed of those who defend that robbery (whether they benefit from it or not). But I don't get into second order beneficiaries. Is a public health doctor net negative? Is a university professor? Where's the line?

2026-06-19 13:14:45 UTC

- reply

Monero is not (completely) private. And doesn't scale well.

2026-06-16 20:53:58 UTC

How would Bitcoin work if there was a human settlement on Mars? And further away?
The time it takes for electromagnetic waves traveling at the speed of light though open space is between 3 and 22 minutes depending on the relative position between the Earth and Mars.
So for example, Mars miners would be at a great disadvantage at first! And reorgs (chain forks) would occur more often.
For regular transactions, it would just mean higher waiting times for Mars users.
But Lightning Network path finding... Probably impractical!
What if there's a human (or another intelligent life form) settlement further away? As in weeks or light years far?
Should each 10 light minute radius area get their own coin? And then the market does its thing setting the exchange rates? Or?

2026-06-16 19:22:24 UTC

- reply

That's called a private company 🤔

2026-06-16 18:53:31 UTC

- reply

how do I buy musk without kyc

2026-06-16 12:53:59 UTC

- reply

timechain!

2026-06-14 19:31:26 UTC

- reply

ack

2026-06-12 09:29:45 UTC

https://bitcoincore.org/en/2026/06/06/privatebroadcast-ip-leak/

2026-06-11 20:33:50 UTC

- reply

Is that so?

2026-06-10 21:03:03 UTC

- reply

Woild be ao mice though :(

2026-06-10 20:40:46 UTC

- reply

congrats @nprofile…pt0r ;-)

2026-06-10 20:33:24 UTC

Stablecoins are winning. We can use this as an opportunity.
Slides from @nprofile…pptm
https://blossom.primal.net/6a3b3af1b5397ae2a2682c0266ec01f60e30a4df50026551b716b6abee9d5798.jpg

2026-06-10 16:14:10 UTC

complexipy: Code complexity analyzer
https://github.com/rohaquinlop/complexipy
Blazingly fast cognitive complexity analysis for Python, written in Rust.
> Cognitive complexity measures how hard code is to understand by humans, not machines.
Unlike traditional metrics like cyclomatic complexity, cognitive complexity accounts for nesting depth and control flow patterns that affect human comprehension. Inspired by [G. Ann Campbell's research](https://www.sonarsource.com/resources/cognitive-complexity/) at SonarSource, complexipy provides a fast, accurate implementation for Python.
**Key benefits:**
* **Human-focused** - Penalizes nesting, flow breaks, and human-unfriendly logic
* **Actionable insights** - Identifies genuinely hard-to-maintain code
* **Different from cyclomatic** - Measures readability while cyclomatic measures structural, testing, and branch density
Try it with:
```shell
pip install complexipy
complexipy . --failed --suggest-refactors # in a python project
```
Example output for JoinMarket NG:
```
joinmarket-ng/tumbler/src/tumbler/runner.py
TumbleRunner::_run_taker_phase 19 ❌ FAILED
Refactor plans:
1. Extract complex block into helper function, lines 351-405
estimated: 19 -> 7 (-12)
steps:
- move the selected block into a private helper
- pass required values as parameters
- return the result/status needed by the caller
TumbleRunner::_run_maker_phase 24 ❌ FAILED
Refactor plans:
1. Extract complex block into helper function, lines 496-523
estimated: 24 -> 6 (-18)
steps:
- move the selected block into a private helper
- pass required values as parameters
- return the result/status needed by the caller
2. Flatten loop body with continue guards, lines 500-521
estimated: 24 -> 14 (-10)
steps:
- add negative condition checks at the top of the loop
- use continue for skipped items
- keep the main processing path unindented
3. Flatten nested condition block with guard clauses, lines 505-506
estimated: 24 -> 22 (-2)
steps:
- invert the outer condition
- return early when the condition fails
- move the main success path one indentation level left
- repeat for inner nested conditions where safe
```
https://stacker.news/items/1505973

2026-06-10 11:53:16 UTC

- reply

Cyberpunks write code.

2026-06-10 11:45:40 UTC

- reply

crazy

2026-06-10 11:40:11 UTC

- reply

nice try NSA

2026-06-09 13:06:37 UTC

- reply

Oh! I really like that idea.
I guess humans having something to lose is also what makes us "lazy". We cut corners and make mistakes because we try to save time.
It is really amazing what one can achieve with the non laziness properties of AI. I've tried many times to have my papers/books directory organized (directory structure, filenames, metadata) and failed. It just takes to be in a rush once to open the door to chaos.
But with AI, I just throw new stuff under pending/ and trigger the AI to follow the recipe. And it changes completely the mental cost of adding new content to the library.
Another big advantage is probably preventing writing fingerprinting. For anyone that has written enough publicly, there's probably sufficient information to uniquely identify that writing style.
Interesting.

2026-06-09 12:57:18 UTC

- reply

Cashu is not Bitcoin. It's debt. An IOU.
But financial debt is not bad on its own. The problem is usually the debtor. Fiat in a bank is debt squared. With the additional disadvantage of 0 privacy and collaboration with the state.

2026-06-09 12:22:25 UTC

- reply

What if Bitcoin already had all it needs to scale and do so with decent privacy?
I want my friends&family Bitcoin services to be as good as my Jellyfin+Ombi setup. People *ask* for an account, it's not me pushing it to them. Because it's better than Netflix, easy to use, and it just works.
Now, I want to improve my cashu mint setup. Find a nice mobile wallet to recommend (looks like minibits for now), and have a fast and robust LN node underneath so that payments work and settle fast.
Will share the final setup once it proves to work for my users.
#naddr1qq…4gxm

2026-06-09 07:55:21 UTC

- reply

#nevent1q…565g
The idea of the next Satoshi Nakamoto being non-human left me thinking last night. I think that won't be the case, because they lack something: caring. They just don't care. They have nothing to lose!
We could already throw an AI into a VPS with a LN wallet with some funds and ask it to run a service or something to eventually "make a living" and pay for its tokens. But why would it? Any mistake that kills its "cron" or main process would "kill" it. And I'm super sure that it will make such a mistake rather quickly. Because it doesn't care.
Even if we got through that, imagine it stays "alive" for long and creates an online service that people pay for (idk, something like nadanada/LNVPN, or nostr.wine). Does it really care about customers? It would be great at apologizing, but probably not enough to keep the service reliable and prevent mistakes on the first place.
I was thinking about this for JoinMarket development: couldn't an AI just handle questions, issues, decide the roadmap, and work on the code? Technically yes. But it probably would take at most a few weeks until there's a funds loss bug or the software becomes unusable due to feature bloat for example.
The difference is that a human cares. The same thing that makes us reluctant to admit mistakes, is what motivates us to avoid them "harder". Ego? Reputation? Responsibility? An AI doesn't have that, for good and for bad.
That's why AI produces sloppypasta: #naddr1qv…nsa5

2026-06-08 19:37:37 UTC

Pieter Wuille (fun) Facts
http://pieterwuillefacts.com/
https://stacker.news/items/1504789

2026-06-08 18:42:21 UTC

- reply

sad :(

2026-06-08 15:14:17 UTC

Am I Being Pwned? - Browser Extension Scanner
https://amibeingpwned.com/
https://stacker.news/items/1504617

2026-06-08 14:51:15 UTC

DeepSWE: Measuring frontier coding agents on original, long-horizon tasks
https://deepswe.datacurve.ai/
![benchmark](https://m.stacker.news/144105)
> Today's leading public coding benchmarks are starting to saturate at the frontier: top models cluster within a narrow score band where adjacent configurations often overlap on confidence intervals. DeepSWE is a long-horizon software engineering benchmark built to separate them. It delivers four advances over existing public benchmarks:
> **Contamination free**: Tasks are written from scratch, not adapted from existing commits or PRs, so no model has seen the solution during pretraining.
> **High diversity**: Tasks span a broad pool of 91 repositories across 5 languages.
> **Real-world complexity**: Prompts are ~half the length of SWE-bench Pro's, yet solutions require 5.5x more code and \~2x more output tokens.
> **Reliable verification**: Verifiers are hand-written to test software behavior rather than implementation details.
> The result is a benchmark that reflects how today's frontier coding agents actually perform in software engineering work.
https://stacker.news/items/1504607

2026-06-08 14:28:05 UTC

How Did DeepSeek V4 Make V4 So Cheap? - YouTube
https://www.youtube.com/watch?v=gC76aeibdFA
Paper: https://www.alphaxiv.org/abs/deepseek-v4
https://stacker.news/items/1504595

2026-06-08 08:15:55 UTC

Looking for feedback on https://github.com/joinmarket-ng/jmp/pull/3 (Taproot offers)

2026-06-08 06:04:56 UTC

The Economics of Protocol Design: Proposing Lightning over Ark‬ - YouTube
https://www.youtube.com/watch?v=hqKQ4Rky3tY
René Pickhardt at BTC++ conference.
Description: Venture Capital Money has a surprising influence over protocol design. Not rarely technical sound solutions won't find consensus due to market sentiment and selfish interests of participants. We investigate this with the example of the debate about ark vs. lighting network as the future payment system for Bitcoin.
Paper: [https://arxiv.org/abs/2601.04835](https://arxiv.org/abs/2601.04835)
https://stacker.news/items/1504446

2026-06-07 19:57:10 UTC

- reply

yes! what's about to say that. And that did not save them

2026-06-06 19:21:33 UTC

- reply

And if you set the minimum reserve to match those, there's no security problem anymore.

2026-06-06 19:19:31 UTC

- reply

It uses the underlying node default fee limits. Right?

2026-06-05 10:29:44 UTC

We're now co-signing JoinMarket NG releases!
@nprofile…79aj has reviewed up to 0.30.0 and that's the first co-signed release.
https://github.com/joinmarket-ng/joinmarket-ng/commit/9a6e8d4333ff716336c05671e07bd9454dc7f43d
He also fixed a docker image reproducibility issue.
Still looking for more eyes. If you can review the code and want to co-sign with us, please reach out.

2026-06-05 08:40:32 UTC

- reply

do they? they look more like inflation "detection" more than explanations on why it would be unlikely for it to happen in the first place

2026-06-05 08:21:34 UTC

- reply

What's interesting is how they manage to keep the TRY exchange rate more or less stable against other currencies. Prices have rosen like crazy but the TRY/EUR pair has barely moved.

2026-06-05 05:44:04 UTC

- reply

That doesn't solve the "censoring". Primal uses a special caching relay for reading. Which is not a nostr relay but some custom thing. I guess you could self host that, but it's not as simple as configuring additional relays.
On top of that they have an npub blacklist which I'm not sure where it comes from and that might not be easy to avoid.

2026-06-04 20:49:48 UTC

I have a problem. I only like Primal for its UX but I'm getting tired of "mentioned event not found". Did not come to nostr for censorship but for censorship resistance.
Any Nostr client recommendations? Have tried Iris and wisp already and was not convinced :(

2026-06-04 18:10:55 UTC

- reply

Good luck enforcing that🤭

2026-06-04 06:20:36 UTC

- reply

sad :(

2026-06-02 08:18:08 UTC

- reply

sad :(

2026-06-02 08:14:14 UTC

Fast, efficient, and high-quality text-to-speech for GrapheneOS
https://github.com/GrapheneOS/SpeechServices
Speech Services provides fast, efficient, and high-quality text-to-speech for [GrapheneOS](https://grapheneos.org) using state-of-the-art models running completely on-device.
https://stacker.news/items/1500823

2026-06-02 08:10:07 UTC

- reply

The relay is not the problem. There are many relays already. But Primal is designed around their caching relay which is not easily replaceable.

2026-05-31 19:16:20 UTC

- reply

Sure. But there's a big difference between who has access to it and what it stores. Selectively excluding notes sounds pretty intentional.

2026-05-31 11:40:26 UTC

Is Primal censoring stuff?
https://stacker.news/items/1499789

2026-05-31 11:23:24 UTC

- reply

2026-05-30 20:48:23 UTC

Cashu Mint Auditor
https://audit.8333.space/
This app sends random Lightning payments between mints. Forever. The sats go in a circle until all is lost to fees or rugs. If a mint fails to pay an invoice, it is marked with a scary red label.
https://stacker.news/items/1499492

2026-05-30 15:43:07 UTC

- reply

can't zap you! but thanks :-)

2026-05-30 15:01:28 UTC

- reply

do you have a link to it?

2026-05-30 12:50:59 UTC

Wouldn't it be nice so receive Silent Payments and CoinJoin them after?
Or how else could a developer receive Bitcoin donations with some privacy?
PR for Silent Payments (BIP352) support for JoinMarket NG: https://github.com/joinmarket-ng/joinmarket-ng/pull/513

2026-05-30 11:39:17 UTC

- reply

Have been using the now available Android app for a while now and unfortunately will have to go back to WireGuard configs. Mainly the missing features are:
- Widget to temporarily toggle it off. And that still works with the "Always on" VPN setting. So that you can actually use the Internet connection with the VPN temporarily off through the widget.
- Exclude addresses on local networks.
Other things are not that important. But can't live without these two!
Super happy with the service in general though.

2026-05-30 11:15:16 UTC

- reply

How does the Lightning channel management work? Is the Bitcoin node within the TEE or external?

2026-05-30 10:55:57 UTC

Privacy-preserving license plates using blind signatures
License plates are crazy from a privacy standpoint. We'd riot if the government made us wear a badge with our name on it every time we left the house. But bolting a permanent unique identifier to our car? Totally normal.
Now there are ALPR (automatic license plate reader) networks everywhere, public and private, logging everything. Check https://deflock.org/ for the USA picture. From plate data + camera feeds (public or insecure private ones) you can infer where someone lives, works, when they're not home, who they meet. Some countries even let you query the vehicle registry directly from the plate.
I don't think accountability for cars is necessarily wrong. If you rear-end someone and flee, it's fair they can identify you. I just don't think we need to trade mass surveillance for that. Cryptography exists.
A basic scheme:
Every day you generate a random string r, compute H(r), and get it blind-signed by an issuing authority (they sign without learning what they're signing: Chaumian blind signatures, same as ecash). You put r on your e-ink plate as a QR code, along with the signature. Anyone can verify: hash the preimage, check the signature. If you cause an accident, the witness notes r and can prove they saw your plate. To find out who you are, they go to court, the authority looks up who requested the blind signature for H(r). No public link between your plate and identity. Already way better than today.
Problems and improvements:
Daily rotation is too slow: Cameras still build a full-day profile. Rotate every 5–10 minutes instead. Eric Rescorla arrived at a similar scheme: https://educatedguesswork.org/posts/license-plates/ and explores the tradeoffs well, including how the authority can precompute a lookup table for all vehicles and time windows cheaply enough that de-anonymization doesn't require your cooperation.
The authority is still a chokepoint: Whoever registers you and creates the H(r) -> identity link already knows both sides. Threshold encryption of the stored record is a governance improvement (requires k-of-n parties to cooperate, auditable) but not a fundamental one. The registrar still has the information at creation time. The real protection is legal process + the fact that passive surveillance is broken. That's already a big deal compared to today, but it's not zero-trust.
The cryptography is production-ready. The gap between "solved" and "deployed" is almost entirely political.
Posted to https://stacker.news/items/1499201

2026-05-30 10:10:38 UTC

Rewrote the "JoinMarket Maker Clustering and Taker Anonymity-Set Reduction" article with a clear example as introduction, a glossary, and trimming most sections.
Published at https://joinmarket-ng.github.io/coinjoin-simulator/mainnet-deanon.html (same URL, old version preserved in git).
The discussion about possible mitigations is ongoing at https://github.com/joinmarket-ng/joinmarket-ng/issues/508
BTW the JM protocol is not broken, just improvable.

2026-05-29 15:27:47 UTC

JoinMarket NG 0.31.0 release is out.
More improvements about deep wallet scanning. Also security hardening, TUI improvements, and bug fixes. Automated Win/Mac/Linux install tests.
https://github.com/joinmarket-ng/joinmarket-ng/releases/tag/0.31.0

2026-05-26 13:38:14 UTC

Max Hillebrand: The Praxeology of Privacy - YouTube
https://www.youtube.com/watch?v=A4_1E2kJ5_w
Timothy Allen speaks with Max Hillebrand.
Max joins me to explore freedom, privacy, and property rights in the free market of ideas. We discuss Cypherpunk strategies, the appeal of Bitcoin, Free Cities, and digital nomadism as ways to resist creeping socialism and surveillance. Drawing on Lockean ownership theory, Austrian economics, and praxeology, Max makes the case for combining decentralization, community, and technology to defend liberty in both physical and digital realms.
TIMESTAMPS:
0:00:00 — Coming up
0:00:41 — Veritas Village preview
0:01:27 — Start of conversation
0:10:02 — Triangular Interventions of the State
0:11:26 — The banning of VPNs
0:16:27 — Mean Time To Harassment
0:24:43 — The Cypherpunk Ideal: Increasing the Cost of Attack
0:30:38 — Dragnet Surveillance & Privacy
0:36:48 — Free Cities and The Importance of Freedom in Meatspace
0:44:04 — There are No Frontiers Left
0:53:08 — Conscription is Coming Back
1:00:57 — There are Many More Good People Than Bad
1:06:53 — AI and Robots of Convenience in the Dystopian Future
1:18:25 — Bitcoin Proves John Locke's Theory of Property Rights
1:25:29 — Proving Economic Reasons
1:31:40 — The Cypherpunks Don't Know How Fundamentally Correct They Are
1:36:47 — Freedom is Correct. The State is Evil
1:49:22 — AI Will Remove Scarcity from the World
1:53:50 — Keynsian Bullshit
1:58:08 — Tik Tok Will Eventually Generate 100% AI Content
2:10:35 — Rally Cry for Freedom Lovers
NOSTR:
Max Hillebrand: npub1klkk3vrzme455yh9rl2jshq7rc8dpegj3ndf82c3ks2sk40dxt7qulx3vt
Timothy Allen: Search '[[email protected]](mailto:[email protected])' on your Nostr app
Free Cities Foundation: npub1lsj8pmgedqqamt89c27tzjjnlf0wn7q7udjm7j2cl9xxz97eacns2mwpee
https://stacker.news/items/1496628

2026-05-26 13:34:52 UTC

- reply

Please no! xD
Take everything as open for discussion, and expect only optional changes that users will decide if they use or not.

2026-05-26 13:15:57 UTC

- reply

Well you don't need the lightning node to run within the enclave right? Just a "smart" signer that signs HTLCs given the redeemed cashu token. For rebalancing maybe some rule of "not loosing funds" and someone external has to pay the rebalancing fees. You can prove to the enclave that the rebalancing succeed or that a LN invoice for which it only revealed the payment hash has been paid.
I'm sure there are lots of challenges here, but the main point is that it doesn't need to run "everything".

2026-05-25 21:02:04 UTC

Been digging into JoinMarket maker clustering on mainnet. the short version: yes you can cluster makers by fee fingerprint from onchain data alone, and yes it reduces taker anonymity sets. But JM holds up pretty well in practice. mean anonset goes 7.6 -> 6.9, and the mitigation (fee policy homogenization) *can* be a client default change, no protocol surgery needed.
The counterintuitive part: some "obvious" countermeasures like makers avoiding change as input actually make things worse.
Rough draft, not peer reviewed, happy to get feedback: https://joinmarket-ng.github.io/coinjoin-simulator/mainnet-deanon.html