A 'Senior Principle Engineer' in my fiat job thinks that the role of AI is to infer the spec from our existing codebases (they are unmaintainable, underspecified, untested, human slop), so that every time a new 'vastly superior' model appears we should ask it to rewrite the code from scratch using this inferred spec as the guide #NGMI "... extract a spec (product, engineering) from an existing codebase, force an isolated AI agent to rebuild it from scratch, and measure the divergence. The goal is to discover the minimal codebase practices we must adopt today so that when vastly superior models release tomorrow, we can confidently hit “re-generate” on entirely AI-generated codebases and guarantee the output is strictly better."

Trying to ban myself from working on freedom tech today, as I've been neglecting my fiat-mining obligations I've been obsessed by my open source stuff, because it's much more fun and important, but I need to do other boring stuff to pay the bills

I think I finally fully understand Ark, and will try to explain it here: (The swapping from the old round to the new round was what I didn't fully understand until I asked AI a few questions this morning) Lightning already shows how two users can share one UTXO; both parties have a pre-signed transaction that they can broadcast at any time to spend the onchain 2-of-2 UTXO. That transaction spends that one UTXO and creates two new UTXOs which give each party their balance. In Ark, it's basically the same, but with any number of users. The onchain UTXO is an n-of-n multisig instead of just 2-of-2. All parties have the necessary transactions which they can broadcast at any time. So every user has 'unilateral exit', they can broadcast these transactions at any time and get a single onchain UTXO with their balance belonging just the them. These are the VTXOs (Virtual UTXOs). You can exercise your unilateral exit at any time by broadcasting your VTXOs. (Pedantic detail that you don't really need to know: if there are 5 users, it's not a single transactions that spends the 5-of-5 *directly* to five different UTXOS, one for each of the five users. There is a tree of transactions, and each user has a copy of the *chain* of transactions that are necessary to get their balance out to a single UTXO under their control) It's possible to receive payments at any time from another Ark user. But it requires a bit of trust (because the Ark Service Provider [ASP] and the sender could collude to take payment money that they sent to you) and therefore I won't discuss it in detail. Ark doesn't use the cool Lightning tech for instant settlement with commitment transactions. Therefore, I'll ignore this slightly-trusted part of Ark, and I will focus on the fully trustless parts, i.e. where Ark users have unilateral exit If you already have 1000 sats locked up safely in one 'round' onchain, and you receive another 500 sats and you want those to be locked up safely, then you will need to take part in another round. This is where things get really interesting, and this is the bit I only just fully understood today ... The ASP creates a new 'round', i.e. a new multsig output onchain that includes all the users who wish to take part. In this round, you will be assigned 1500 sats. i.e. you now have unilateral exit allowing you to take 1500 sats But you still have the VTXOs from the first round, and so it seems like you could implement both exits and thereby claim 1000+1500 = 2500 sats, stealing 1000 sats from the Ark Service Provider The solution is a 'forfeit' transaction. The ASP won't include the 1000 sats in the new round unless you first sign a 'forfeit' transaction and give that signature to the ASP. This gives control of your original 1000 sats to the ASP; if you broadcast the original VTXOs, then the ASP can use the forfeit to claim those 1000 sats for itself. There is a timelock in the VTXOs, in order to give the ASP sufficient time to broadcast the forfeit transaction The forfeit therefore protects the ASP, because it stops the user from 'double exiting'. There can then be multiple on chain 'rounds', and you have a balance in all of them, but you are practically able to access only one of them [the latest one]. 'Your' money in the older round UTXOs isn't really yours any more. But we must also protect the user! You, the user, don't want to forfeit your balance in the 'old' onchain UTXOs before the new 'round' is confirmed on chain. The solution here is 'connector outputs'; the forfeit transaction depends on outputs created in the new round and therefore the forfeit is ineffective until the new round is confirmed. So now both the user and the ASP are protected. The user has a balance on-chain in the 'old' round, with full unilateral exit via their VTXOs, and they can trustlessly and atomically 'move' that into a 'new' onchain round. It's not instant, because the new rounds need to be confirmed on chain, but it is unilateral exit So Ark is great at scalability, but it has a few downsides for the various parties: It's not private If you insist on full sovereignty, i.e. you don't trust the ASP, then Ark is slow like on-chain. If you want instant trustless final settlement, then Lightning is still the only option From the ASP's point of view, they have to lock up a lot of liquidity. If a given user has a balance of 21,000 sats and there are five rounds with that balance, then the ASP actually has to lock up 105,000 sats on chain. Each round therefore has an expiry time built in, where all the funds go back to the ASP after that expiry time; this motivates all users to take part in a new round - or exit or otherwise spend their funds - before the expiry. This sounds like I'm being quite critical of Ark. It sounds like I'm saying that it's as slow and as non-private as on-chain, and the only advantage is scale because many users can share a single UTXO. However, this is where Lightning comes in again. Lightning shows how you can take a slow and transparent system, and then build private instant settlement on top of it In the vision of 'Ark as a channel factory for Lightning' that René Pickhardt and others have been talking about for a couple of years, everyone can have a full Lightning node. They haven't channels with other people, doing (blind) routing over the Lightning network. They'll be paying for preimages in the usual way. They'll be sending 'commit transactions' over and back in the usual way. Fully Lightning in every way, no compromises. In this vision, everyone gets private instant settlment because they directly use Lightning The subtlety in this approach is that, instead of Lightning channels being backed by conventional already-confirmed on-chain funding transactions, the funding transaction that backs the Lightning channel is instead an Ark VTXO. You and I can set up a Lightning channel without broadcasting that funding transaction; we are both satisfied that either of us can - at any time - broadcast the Ark VTXOs to get the funding transaction on chain. This allows us to do all the Lightning stuff *as if* the funding transaction is already on chain. In this context, you would have to 'exit' twice in order to get your funds. The first 'exit' is the Ark exit, where you broadcast the VTXOs in order to create the 2-of-2 funding transaction for the Lightning channel, and then the second exit is where you broadcast your latest Lightning commitment transaction in order to split the the 2-of-2 into the final on chain balance for you and for your channel partner In the 'channel factory', Lightning is still the Layer 2. All payments fully go through Lightning, giving us private instant settlement. Ark is like 'Layer 1.1', offering a wrapper over Layer 1 that is still slow and somewhat transparent, but which allows us to scale up the number of users that can fit on chain

The British sense of fun around elections is great. This is a genuine photo from last night of the UK's likely next prime minister with other candidates for the seat that he has just won. In every election, fun candidates run against the high profile politicians

I ran twice today. 7.5km each time It's nice to suddenly be addicted to jogging again

I really want to buy some STRC now, it's just $86. Partly to troll the maxis, just buy a small amount for fun But I bought too much BTC already this month and can't afford it!

On Twitter, I'm starting to unfollow people that give too much attention to BIP-110 I generally enjoy the content of folks like @npub1xapj...cnde and @npub1qg8j...24kw , but BIP-110 is now so boring to me that I unfollow them until this is all over and until they stop giving oxygen to BIP-110 and to unrealistic relay policy generally The "debate" died long ago, and we (the tolerant minority) have won Even the Bugle @Bugle.News #40HPW🎧 are running out of funny things to say about it; surely a sign that this topic is long dead I first got properly into Bitcoin only last summer, and I had great fun arguing with Knotzis as I learned more about the topic. But there is nothing to be gained any more. Shinobi, it's time to accept that you won't get any satisfaction from Hodlonaut I'm not sure if I'll engage with any replies; but of course I'll engage in good faith if I have met you in person and know that you're in good faith

Great story View quoted note →

I'll go to Bitdevs tomorrow. I missed it the last few months, sometimes from laziness and sometimes because I was out of town Looking forward to it!

The next interesting step for MONAD (Tor/VPN, funded by Cashu Spilman channels) is adding a nice management API; then I hope to have a nice dashboard web-app (maybe a PWA) to monitor and manage the wallet and the client and the relay. I've been adding a lot of smaller things recently, like blinded paths and more stress testing and other little improvements. But it's time to wrap this up into something easy to use, like a Docker container with easy configuration and monitoring. Then I can start asking people to do some testing View quoted note →

Stop pretending that the UK's problems around internet freedom are because it's "left wing" or "communist" or "woke" The current UK Labour party are famously quite right wing compared to previous Labour governments, and the UK generally is a bit more right wing than many European countries such as France This isn't a left versus right issue. It's a mixture of (Epstein) evil, and incompetence The manipulators want you to think that it's a left versus right issue, so that they can distract you with culture war crap. Engage your brain, and keep your eye on our real goals

I keep trying to do my fiat job, but I keep getting drawn into freedom tech instead ----- I've been busy with MONAD (cashu-funded Tor) in recent weeks. Various protocol improvements, and general stress testing under hundreds of simultaneous connections and hammering the payment channels I started using KImiK2.7 and I like it so far. I'll use it for at least a few more days before trying to make a final judgement on it

reading Montaigne's essays

A longer-term idea to minimize prompt injection: Hopefully, the future will be all open models and they'll all be running on commodity server farms (using something like @routstr to buy and sell) I know there is randomness when predicting tokens, but we should make it deterministic, i.e. pseudorandom Each server should sign every response. We should send a small fraction of our requests to two servers simultaneously. If two servers give a different response to the same request, then it's likely one has cheated. With the signatures, we can then prove which server farm misled us I guess this is a bit complex, but it's a fun problem to think about View quoted note →

Good read, on how Liquid keeps amounts private, while also being immune to the kind of issue that affected zcash recently (Not very much technical detail though; I should follow the references here to learn more) View quoted note →

BTC Yield will continue until morale improves View quoted note →

I'm reading a Culture novel again now, for the first time since getting properly into Bitcoin and AI There is a lot of AI in those stories, so it's interesting to read it now that it seems more realistic And they're always pointing out that the Culture isn't so "organized"; no hierarchies and laws and so on. I guess it's what some anarchists would hope is possible (They still have extraordinarily powerful armed forces though)