01:09:22 Hi — anonymous dev here. I'm launching a RandomX CPU-mined L1 (fair launch, no premine, no admin keys) whose design deliberately borrows from and diverges from Monero's, and I'd rather be attacked by this room than by nobody. 01:09:43 https://github.com/thesimstoshi/zycord 01:09:56 The core idea: each transaction is a certificate that declares the state it read and the state it wrote, and carries the proof. Verifying it is therefore a pure function of its bytes — no database, no history, no synchronisation — so verification parallelises without bound. The chain never executes anything; it orders cert [... too long, see https://mrelay.p2pool.observer/e/8YrA8qYLdU5tYlJh ] 01:09:56 That skip mechanism is what makes two fee markets possible from genesis: sequential gas for state mutation (scarce), parallel gas for verification (abundant — it grows with every core that joins the network). Both are EIP-1559-shaped. Skipped certificates burn their fee and pay nobody, so a block producer maximising revenue [... too long, see https://mrelay.p2pool.observer/e/8YrA8qYLdU5tYlJh ] 03:32:14 thesimstoshi Use RandomX v2, for starters 03:55:43 @thesimstoshi:matrix.org Commit messages make the project seem AI-generated. https://github.com/thesimstoshi/zycord/commit/c861966a90ef3edf373047f208e5df931b754996 is a good example. 03:56:04 Who write these monster commit messages for such minor changes? 03:56:41 Mucho em dashes 03:56:55 Using Claude? 03:57:23 I'm not as familiar with Claude's writing style, I can't tell which model it is. 03:57:54 @torir:unredacted.org: Im pretty sure it’s claude, not sure about the exact model. 03:57:54 Claude often do that kind of stuff 03:58:23 @thesimstoshi:matrix.org: Try to lobotomize your AI agent, you’re wasting money and time in commentzilla 04:01:06 @slowbeardigger:matrix.org: I meant the Claude family of models vs other model families. 04:04:43 I'm not going to tell you to not use AI, but I'd respect the project more if AI use was disclosed (co-authored-by line in commit messages would be enough), and if you at least showed some sign of having a human clean up the commits (for instance, removing the unnecessary and lengthy comments). 18:36:47 Yes, I'm using Claude quite a lot for programming. I developed a complete version of the code by hand to validate the concept internally. But it became huge and had many bugs, so I made this second version with AI and stayed on top of it, making the decisions and managing the agents. 18:36:47 This second version turned out much more robust, even with AI. 18:37:39 I'm using Fable 5 and Opus 5. 18:45:38 So, I'm not really looking to make money off this project. I've been burning more cash on it than anything else. That's why I'm trying to validate the idea here. I've actually been at this for the last ten years. I've poured a ton of money into making this project a reality, always kicking it down the road because I felt it wa [... too long, see https://mrelay.p2pool.observer/e/iP_ZkKcLQTBNRmFC ] 18:49:36 Thanks for the feedback, I'll do that. > <@torir:unredacted.org> I'm not going to tell you to not use AI, but I'd respect the project more if AI use was disclosed (co-authored-by line in commit messages would be enough), and if you at least showed some sign of having a human clean up the commits (for instance, removing the unnecessary and lengthy comments). 19:12:32 If the central thesis is wrong, I'd rather find out now than spend more time and money. At some point, I need to come out of my cave. 19:13:25 You are going to get a lot less feedback on anything AI generated. I bounced right off your proposal and only reply now that you are typing stuff yourself (or so it seems). 19:19:09 Put the AI aside for a moment. I did what I could with what I had. I'm a solo dev. Years ago I tried to get this idea off the ground, but without resources it's really hard. Besides, I'm not much into fundraising. 19:28:16 The design's big bet is a division: everything that can be independently verified runs in parallel, leaving sequential only what is genuinely sequential. I drew inspiration from my studies with Hyperledger and other networks. Allowing transactions to be validated by a GPU. Only the ordering of states/balances becomes sequential. No execution. Trivially verifiable. 19:30:51 The trade-off is that transactions become highly complex at the construction phase, requiring a Web2 infrastructure to be set up to ensure services run properly. However, the gain is enormous as well: massive parallelization. In the future, when mining migrates to PoS, each node won't need to validate everything. If each valid [... too long, see https://mrelay.p2pool.observer/e/jNv_kacLcGI1N0Fp ] 19:31:13 Enabling complex applications, not just a CryptoKitties. Things that today only run on Web2 could run on Web3. 19:31:25 Am I delusional? Am I crazy? 19:32:11 "in the future when", when whaT? 19:32:12 lol 19:32:27 we are migrating from randomxv1 to randomxv2 19:56:16 Nice, I'm going to prepare the migration to use RandomXv2 on testnet. It seems to be quite simple. 20:02:41 There is a pr for that already iirc 20:03:36 10032