<unnamed>

Public Key

npub16zsllwrkrwt5emz2805vhjewj6nsjrw0ge0latyrn2jv5gxf5k0q5l92l7

Profile Code

nprofile1qqsdpg0lhpmph96va39rh6xtevhfdfcfph85vhl74jpe4fx2yry6t8spp4mhxue69uhkummn9ekx7mqpr9mhxue69uhhq7tjv9kkjepwve5kzar2v9nzucm0d5q3gamnwvaz7tmjv4kxz7fwv3sk6atn9e5k7qgkwaehxw309ahx7um5wghxcmmjv4h8g73wd9esz9thwden5te0wfjkccte9ehx7um5wghxuet5qydhwumn8ghj7emvv4shxmmwv96x7u3wv3jhvtmjv4kxz7gmr5wn2

Publishing to

gleasonator.dev/relay

Last Notes

2026-09-09 19:09:24 CEST

- reply

Yeah the encryption is what protects the raw share data from the transport infrastructure. Each key (Shamir share) is gift wrapped to the intended steward's private key. The invitation link dance when inviting stewards (which does use a horcrux url, but doesn't have to) is essentially setting up an authorized set of keys so when your app sees a recovery request it's easy to discard bad actors.
So Nostr is a peer to peer transport layer but it's also the identity layer. The nostr events only need to stay on the relay long enough for the stewards to fetch the latest ones, so in most cases just a few days. And yeah I think being able to choose your own server is a big deal for something you want to last a long time. If I had used p2panda or iroh or ssb instead of Nostr I would still need a store-and-forward server.

2026-09-09 18:53:44 CEST

@npub107j…ncxg I've been using fragua for a couple weeks now and I quite like it. Some feature requests:
- i want tool steps to have structured output (for example i have a tool to compute branch name). Right now only llm steps can provide structured output? I've hacked around this by having an llm call my tool, but it's not ideal.
- I wish the webui had some more contextual actions (i.e. approve/decline changes) on the run's detail page
- I wish there were more timestamps in the run list and detail pages.
- I want some control over worktree creation. I have a workflow that I use for feature work or if I pass it a PR url it looks at my review and addresses it. This leads to worktree collision where sometimes the initial run for that branch left tree dirty or the branch checked out and fragua fails to account for this. I've worked around it temporarily by creating a new branch for every run.
I'm not asking you to implement all this. Just letting you know in case it dovetails with feedback from other users. I am not using it as intended where I want to apply the changes to my local git worktree at the end of a run. I am running fragua on a remote sandboxed machine and most of my workflows open or update a PR that I review elsewhere. Which makes the Approve/Discard step at the end of the run kind of unecessary. I'm thinking about building my own webui that is beads-aware but still uses fragua to drive workflows under the hood.

2026-09-08 21:15:50 CEST

- reply

Hah. I don't understand. Are you suggesting I just like give people a file and tell them to email it to their stewards? How would the encryption work? I think a big part of making social backup usable is creating assurances that your stewards do in fact still have the data and can access it when needed. You need some reliable machine to machine communication for that. It doesn't have to be Nostr (in fact I am thinking about adding iroh as an alternative transport), but Nostr has the pieces I need and is also built from simple building blocks that will likely still be understandable in 50 years.

2026-09-08 13:17:38 CEST

- reply

Do you have any favorite stories/videos/articles of Bitcoin serving progressive ends?

2026-09-07 22:15:28 CEST

- reply

Now it works (without --fpa) thanks!

2026-09-07 16:18:13 CEST

https://youtu.be/LJKDF0BwepI

2026-09-02 22:18:11 CEST

- reply

CC @npub10np…tl5h auth.nostr1.com seems to only give me events since July, is that the expected behavior?

2026-09-02 22:15:55 CEST

- reply

CC @npub1nlk…jm9c I think nos.lol is having some auth bug? In any case I can't fetch 1059s with nak like: `nak req --prompt-sec -fpa --auth -k 1059 -p @npub16zs…92l7 nos.lol`
I tried with and without --fpa and --auth.

2026-09-02 22:14:44 CEST

What are some good inbox DM relays? I think mine are borked.

2026-08-26 21:11:06 CEST

Just dropped Horcrux 1.0.1 with fixes for a few bugs I found in my final testing before 1.0. Still collecting questions for the FAQ - if you have any ask away. https://horcruxbackup.com

2026-08-21 15:40:52 CEST

- reply

I can't imagine what that must feel like. Praying for you and your family to get home safe.

2026-08-20 17:05:24 CEST

Run `pnpm install`. Some weird error about versions.
Run `pnpm install` again. Different error.
Run `pnpm install` a third time. Magically succeeds.
I hate javascript.

2026-08-20 15:52:22 CEST

- reply

Ah that makes sense. Horcrux couldn't take advantage of that path as it doesn't interact with the blockchain at all. It's much more about local trust. In fact I've been thinking about making it fully p2p because the data model is already there (as it is with most Nostr apps).

2026-08-19 23:07:22 CEST

- reply

The main levers you have to work with are the number of keys you create and the number required to open the vault. So if you are worried about two key holders colluding you can set the required threshold for decryption to 3, or whatever number you want.
It's not a traditional dead man's switch, but it's a tool you could use to set up a social version of the same thing. You'd instruct your key holders not to open until you are dead, and then choose people you trust to honor that. That's exactly my primary use case actually.
I've never heard of decaying key requirements. Is there existing cryptography for that? This is build on Shamir's Secret Sharing algorithm and NIP-59 gift wraps.

2026-08-19 20:13:13 CEST

Today marks the 1.0 release of my app Horcrux 🥳 Horcrux is an app for backing up data like digital wills, passwords, and cryptographic keys. The cool part is that it backs them up to the devices of your friends and family and recovering the data requires their consent. So nobody has the whole secret and even if one or two people lose their piece recovery is still possible. This adds a social layer to your security stack that is highly secure but also highly resilient. You can read more about it at https://horcruxbackup.com.
Of course it's free, open-source, and built on top of Nostr. I'm putting together a FAQ page for the website now, so post your questions in the thread below.
I'd love any feedback you have to give, here or through the contact form on the website. And I'm trying something a bit different. If you are trying to figure out if this tool is right for you and would like to just chat about it you can book a call with me here: https://cal.com/matt-lorentz/horcrux-onboarding

2026-08-19 15:58:47 CEST

I burned several days over the last few months jumping through Apple and Google's App store hoops - DUNS numbers, identity verification, encryption documentation, half a dozen wizards about content in the app, privacy, and of course then the days long review process itself.
I've never published anything to https://zapstore.dev before and I just did it in 20 minutes. That includes me reading through the documentation and futzing with my Android keystore for several minutes which I had previously misconfigured. Absolutely delightful experience.
I'm hopeful that Google's blocking of third party apps and app stores is going to cause a flourishing of third party stores like Zapstore (and de-googled Android). The old systems are broken. I have to assume the deluge of vibe-coded apps is why my Apple review took 18 days. I want an app store where I pay security auditors to compile the apps themselves and go over the code with a fine-toothed comb. F-droid already has some custom repositories that more or less work this way. Zapstore has better infrastructure for this with the Vertex web-of-trust and app catalogs.

2026-08-18 21:51:11 CEST

I submitted the release build of Horcrux to Apple for review on July 31st. Today, 18 days later, I got my first review. But they approved it! So I guess that makes tomorrow launch day 🍾

2026-08-18 21:49:36 CEST

- reply

100%

2026-08-17 21:44:50 CEST

- reply

weird it's always free for me.

2026-08-17 21:44:22 CEST

- reply

I saw that downtime you had 20 minutes ago though

2026-08-16 15:24:01 CEST

- reply

Hi!

2026-08-13 18:46:03 CEST

- reply

ok, I'm trying it.

2026-08-13 18:00:09 CEST

- reply

TLDR; I think your approach is probably superior but prior experience makes me afraid to try it.

2026-08-13 17:35:35 CEST

- reply

Interesting. You don't have a meta workflow shuffling things back and forth between review and fixes? I found myself doing so much busywork shuttling tickets manually between planning -> implementation -> (ai) verification -> (ai) code review -> back to implementation -> back to ai review -> finally to human review, etc.. I really just want to focus deeply on planning something and when I'm done have another high quality thing ready for me to deeply review. I have tried a few DAG things (have you seen Archon?) but they have all been too buggy - agents getting stuck, gates not completing. Or too expensive, like gas town or loa (https://github.com/0xHoneyJar/loa). This drove me towards lighter weight orchestration and having a managing agent that observes and nudges others.
If you don't mind sharing how much does your setup cost per month or per day? I see "12" as the default limit on the review PR, is that $12? I feel quite token constrained atm. While I'd love to fan out 6 lenses in parallel for a PR review I don't think I can afford to do that everyday as a solo dev.

2026-08-13 17:21:05 CEST

- reply

I will get around to this at some point. Already running pomade so what's another key sharding thing.

2026-08-13 17:18:12 CEST

- reply

https://github.com/fiatjaf/nak

2026-08-12 16:31:07 CEST

nak is such a good command line tool it makes me question whether it exists to support the Nostr protocol or if the Nostr protocol is just an excuse to ship nak.

2026-08-07 17:05:27 CEST

- reply

I wrote about the broad strokes in this note: #nevent1q…m7tp

2026-08-07 17:03:24 CEST

- reply

No mine's just my own thing. I've tried many others but none of them quite fit. I'm happy to share the code with you though if you want. I'm not sure the best way to do that but I'll figure something out if you are interested.

2026-08-06 06:15:36 CEST

- reply

Hey glad you are back. Is this something you can share? I have made something similar but I don't have great deterministic gates.

2026-07-29 21:44:39 CEST

- reply

I gave it a short try but didn't find anything that excited me. I'm working mostly solo so the idea of sharing agents with other team members, while really interesting, isn't immediately useful. And they don't have any architecture to enable the type of swarming I'm talking about.

2026-07-29 15:18:14 CEST

I canceled my Cursor subscription today. It has felt for a while that they stopped building for the way I want to work with AI. Then the bugs keep getting worse, and the SpaceX acquisition doesn't give me a lot of hope that things will get better.
I spent yesterday using VSCodium and my homebrew agent orchestration system in tmux. I also tried emdash (https://emdash.ai/), and retried Archon (https://github.com/coleam00/Archon).
I'm still feeling like small agent swarms with a mixture of models that pass work between themselves is the sovereign AI stack of the future. The manager of the swarm can paper over a lot of the small issues with LLMs. Plug in whatever cheap source of tokens you can find, mix in local models if you want, your interface stays the same. Put it all in a decent sandbox and dangerously skip permissions.

2026-07-24 20:51:51 CEST

- reply

Yes we went through a few. Flotilla, zap.stream, routstr, maybe some others. But we actually spent more time discussing some of the tradeoffs of the Nostr protocol than I had planned. Which was nice, maybe I should've organized the session more around that.

2026-07-24 15:33:31 CEST

Has anyone tried https://buzz.xyz yet? Thoughts? How does it compare to something like gas town?

2026-07-23 23:25:58 CEST

I finally finished my security audit of Horcrux, the last major project before release. It was a ton of work and honestly no fun at all, but I'm glad I did it as it gives me the confidence to finally open the app up for real world use. I've got a couple last bugs to fix and some marketing materials to write. Hoping to submit binaries to the app stores next week! If you want to check it out the audit it's published here: https://github.com/mplorentz/horcrux/blob/main/docs/security-audit-july-2026.md

2026-07-22 21:29:36 CEST

The answer, unfortunately, is no.
#nevent1q…e0ul

2026-07-14 22:08:22 CEST

Sweet. I would love a kind 1 client to order comments by one of these algorithms (mainly Most Relevant). I love browsing replies in the reddit-like tree of Nostrudel but often there are spam or off topic comments at the top. @npub1ye5…knpr @npub176p…vgup
#nevent1q…eyvq

2026-07-07 19:55:11 CEST

Any other Nostr apps you would like me to plug?
#nevent1q…vgqr

2026-07-07 14:52:54 CEST

@nprofile…gnen @nprofile…nj27 I just tried creating a new account and using the promenade bunker and got an error when it was trying to "activate the bunker".

2026-07-07 14:52:10 CEST

Is anyone hosting a version of https://use.nsec.app that still works? It seems like it's trying to do some NIP-05 allocation that is broken and I can't skip, and brugeman is gone.

2026-07-01 23:40:45 CEST

I'm hosting a session at DWeb Camp next week for folks to interactively try Nostr. What apps should I encourage people new to the protocol to try? What is the best app to show first?

2026-06-25 22:06:20 CEST

I just published my Q2 report as part of my grant from OpenSats, if you are curious what I have been up to recently. TLDR; getting horcrux into a public beta and preparing for the initial release. https://mattlorentz.com/2026/06/25/opensats-q2-grant-update.html

2026-06-16 22:33:52 CEST

I'm making a push to release version 1.0 of Horcrux this month. Not sure if I will make my deadline but it should be close. There is so much left I want to do in terms of UX, marketing, and sustainability, but it's delivering on its core promise of being a secure way to store secrets with friends and family.
I'm currently in week two of my security audit and I found my first critical issue today. It's humbling but also satisfying to be reviewing all the hard work I've put in over the last year.

2026-06-16 00:26:41 CEST

- reply

TIL about Neo Cypherpunks. More power to them.

2026-06-16 00:19:18 CEST

Pomegranate is another step forward for key management. Glad to see it shipping already!
#nevent1q…94sj

2026-06-13 13:25:44 CEST

- reply

Https://flotilla.social cc @nprofile…dgdk

2026-06-13 13:23:04 CEST

- reply

This is the type of marketing material I want for Nostr.

2026-06-13 03:38:40 CEST

- reply

Looks like you managed to fix it? In any case, it wasn't a big deal for our community. Bugs happen.

2026-06-12 17:25:42 CEST

Claude Fable when I ask it to fix my regex: https://youtu.be/-UYgORr5Qhg?t=55

2026-06-11 14:03:06 CEST

- reply

We are legion @nprofile…ac8d

2026-06-11 13:57:14 CEST

- reply

I went a couple weeks before realizing "silent" and "vibrate" are two different modes. But I had the same experience. I think another problem is that I have trained myself to ignore all vibrations while sitting at my desk because the urgent ones (i.e. wife messages) popped up on my Mac. I also bumped up the time that notifications show on my Pebble to like 5 minutes and that helped as well.

2026-06-11 00:25:52 CEST

- reply

CC @nprofile…8903

2026-06-02 15:43:11 CEST

- reply

Yes but I think Apple like held onto my phone number in the registry or something and it took several weeks before RCS came on in all of my graphene chats. My group messages just went into a black hole and had to be recreated :( if you haven't switched yet I would remove your phone number from iMessage on your iphone several days before you actually switch SIM cards.

2026-06-02 08:49:12 CEST

- reply

https://youtu.be/DYzT-Pk6Ogw

2026-05-30 21:06:26 CEST

- reply

What about key recovery would make you more comfortable with using this? Are you worried about losing access to your files?

2026-05-30 06:58:18 CEST

- reply

Nice, I just did the same. May the RCS spirits be ever in your favor.

2026-05-28 20:27:00 CEST

- reply

Yes

2026-05-22 23:26:14 CEST

- reply

Time to pay for Kagi.com! It's worth it!

2026-05-22 19:42:07 CEST

- reply

What is nunchuck? I haven't heard of that one. I do think the Nostr architecture really fits this use case really well. At first it seems like something that would be better fully p2p, but when you get into the real user needs (non-technical stewards, async recovery, push notifications) having a simple relay architecture with end-to-end encryption is the ticket I think.

2026-05-22 17:40:10 CEST

- reply

The app is designed to be fully tolerant of that! Each steward who holds a chunk of your data stores it on their phone in an encrypted database. You need the relay during the initial backup, during a recovery session, and when changing devices, but the system is designed to sit idle for years (in extreme cases) and still work.
It does suffer from the same problem that most Nostr apps do: if you are only using one relay and that one completely goes away you will have a bad time. Your data is technically safe but recovering it would involve getting all your stewards to (manually) scan some new relay to pick up the new config. I could do some mitigation for that specific case in the future but there's nothing there yet.

2026-05-22 17:34:10 CEST

- reply

Yeah I will definitely add amber/NIP-46 sign-in eventually. Although it's not super high on my list as the app is sort of designed for you to use a fresh npub for it even if you have a Nostr account already. There isn't really anything you get (yet) by logging in with an existing key. I don't read contact lists, profile metdata, etc. to minimize metadata leakage.

2026-05-22 17:32:02 CEST

- reply

Thanks! I can't take credit for that actually. My friend came up with the name :)

2026-05-22 16:17:25 CEST

- reply

Yep! In fact I have a flow that offers exactly that during onboarding.

2026-05-22 15:38:17 CEST

Today I'm opening up Horcrux for public beta testing! If you've ever had a password or document that felt too sensitive to put in a password manager or physical safe, Horcrux is for you. It encrypts and splits your data into pieces and stores it across your friends' devices. Like multi-sig for data. It's free, open-source, and Nostr-based.
I've been testing with a private group as I build but now I'm looking for more folks to help smooth out the experience before people start putting real secrets in it. If you'd be willing to download it and set up a vault with a few friends (or me! ) and put it through the paces I would be so grateful. Let me know how it goes here or using the Send Feedback button in the app.
Android users can grab the APK here (Google Play link pending review): https://github.com/mplorentz/horcrux/releases
iOS users can install via TestFlight from here: https://testflight.apple.com/join/DG5kJpvu

2026-05-22 15:21:34 CEST

༼ つ ◕_◕ ༽つ gib all range-based set reconciliation.
#nevent1q…s3ww

2026-05-22 15:17:43 CEST

- reply

Wow I hadn't heard about about Samsung. I'm hoping that Google's clamp down will just result in more effort going into open source apps, stores, and OSes. Including Linux! More power to you. Switching phones has made me aware of the number of apps I use on a weekly basis, it's a lot more than I thought, and they would be hard to replace.
I don't really believe this yet, but I saw Karpathy say the other day that the next generation of computers will just be LLM interfaces over APIs. Maybe there's a world where I can run a linux phone that is basically just a thin client for my self-hosted Hermes agent running a local model at my house, and it can do all the things the 75 apps on my Android phone currently do 🤷‍♂️

2026-05-21 23:35:34 CEST

- reply

lol ok I scrolled down like one screenful of my feed and saw that you do know Graphene: note1v3zmqfnfa29yksljudyl8rcu7ms3ykwy6nz8346qpnxl8gvzl0asn4zqnj

2026-05-21 23:34:43 CEST

- reply

Have you tried GrapheneOS? I recently made the move from iOS, I've been pretty happy. I know it's not Linux, but it's the least Google you can get on Android.

2026-05-20 14:28:54 CEST

- reply

Unfortunately I don't have a great list of resources. Almost all LLM tutorials/content feels way overhyped and surface level to me. It's all focused on small codebases, and it's extremely hard to find people building normal-sized products for normal users. But I think my entry point was this blog post from Steve Yegge about his beads db: https://steve-yegge.medium.com/beads-best-practices-2db636b9760c (Steve's blog is good but the posts tend to be long and wandering).
More recently I stumbled across this guy on Youtube who seems to be doing some homebrew orchestration, but I haven't watched a lot. https://www.youtube.com/watch?v=PIdETjcXNIk&t=225s

2026-05-19 20:15:33 CEST

I was there today when the Chaos Commoner Collective was born.

2026-05-19 15:21:35 CEST

Against my better judgement I have now homebrewed my own AI agent orchestration harness. I'm running a bunch of different pi.dev harnesses in tmux with a watchdog agent that can restart or send them messages. Each persona (planner, coder, tester, reviewer) has its own prompt and handoff protocol to move work to the next agent. All the tasks and plans are tracked in a beads database. It's a lot like Steve Yegge's gas town but it works much better with my theory that passing work through different models and personas produces much higher quality output. It also works much better with local models and doesn't have a ton of bugs. And it's all running on my old gaming pc with limited permissions to my repos and no credentials for production services.
Since Cursor killed off my previous workflow (pair with a really fast agent to work on a big feature, maybe have a couple easy side quests going elsewhere) I have been trying to switch to a "slow productivity" model of development. By which I mean minimizing context switching and being fully present to review one piece of work at a time. This means I want my AI pipeline to produce the highest quality work even if it takes a long time. And I also don't have a thousand dollars to spend on tokens each month, so I can't just send codex or claude code off to spin for hours on every task. Right now I'm using a combination of GLM 5.1, Deepseek v4 Pro, Deepseek v4 flash, and Qwen 3.6 37B A3B (last one running locally). What are you using for orchestration?

2026-05-16 15:19:13 CEST

- reply

Thank you bot

2026-05-16 15:16:19 CEST

- reply

CC @nprofile…n8v9

2026-05-16 15:15:30 CEST

More of this vibe https://youtu.be/LisPfFa-n7k?si=rOv4DkYrvzJaINbR

2026-05-15 01:47:35 CEST

- reply

Cool, reminds me of https://github.com/Dicklesworthstone/mcp_agent_mail_rust but more focused. I wonder what would happen if you put one agent in charge and have it organize a group of others gas town style?

2026-05-08 18:51:03 CEST

- reply

@npub107j…ncxg is.. gone? 😭😭😭

2026-05-08 18:48:17 CEST

- reply

Good to know. I'm wondering if I had something misconfigured when I tried Qwen 3.6 with hermes because it seems much more responsive now that I'm back using it with pi.dev to write code. I will probably give fully-local-hermes another shot in the future.

2026-05-08 00:06:26 CEST

I haven't shared much about this but in my free time I've been venturing into the self-hosted AI space. I acquired an old gaming machine with a decent graphics card from 4 years ago (RTX 4070S) and put linux on it and spend some time getting hermes agent (https://hermes-agent.nousresearch.com/) running on it.
I got it running with various sparse versions of Qwen 3. Managed to cobble together a few scripts to do things like scrape some news and flight data, but I kept running into timeout errors at various levels of the hermes stack. It's really not set up to work with agents that take multiple minutes to respond and after fixing things in a bunch of different places I got tired of it and switched it back to claude. I did find a fork that supports Zulip and I really love it as an interface for many long-running async conversations.
Then I decided to try to some autonomous coding with local models and fell down hard into the Steve Yegge beads/gas town/gas city rabbit hole. I took gas city (which is like the sdk for agent interactions extracted from gas town) and got it running. I tried running the entire thing with only local models but it wasn't working at all. I ended up with Claude as the mayor of the city who oversees a bunch of short-lived agents that use qwen on my gpu and try to write code and open PRs. They aren't doing a very good job yet but the mayor and I learn and improve things a bit more every day.
I'm not a fan of the super-extractive metaphors of gas town but I do really like beads db as a system of getting agents to cooperate. It's basically an issue tracker, but some issues get labeled as memories and some get labeled as mail and some even represent agents, so it creates an observable system of cooperation where agents spin up, read their mail, complete their task, and hand it off to another agent, then shut down. I'm trying to run all these agents serially to limit gpu contention and it somewhat works. But it's going against the system's design which is just to have a mega bonfire of tokens. The biggest weakness I think is just that the free models that fit in 12GB of vram are not enough to do good coding. But the goal I'm working towards is getting frontier-quality code with free models on my own machine by chaining together enough hill-climing loops (planner, coder, architect review, qa review, bounce it back to coder, etc.) to get good code. And I'm thinking a lot about what the right interface is for me to review the work, right now it's just producing pull requests that I review normally.
This has been my first time running --dangerously-skip-permissions agents 24/7 on my hardware and it feels quite cyborgian.

2026-04-29 16:29:39 CEST

Cursor 3 is so much worse than Cursor 2 that I spent time looking for another IDE yesterday. I thought they really understood how I wanted to work with AI but that trust has been totally shattered. I tried Windsurf but it doesn't have useful git worktree support. I also tried native VSCode with some plugins and Claude Code CLI. All were disappointing and I'm back on Cursor today.
I feel so alone as a Cursor user, I feel like everyone I know is doing CLI development. For me the development bottleneck is reviewing and manually testing AI generated code. Both of these are much easier in an IDE. The UI to approve each hunk of changes the AI made is key for me (after it's done, not the interactive permissions prompts that claude insists on unless you use yolo mode). I need to be able to quickly see more context around the lines the AI changed, click through call hierarchies and go to definition.
Then I want a dashboard that lists all my agents working in different worktrees and I need to be able spawn a new one quickly. And I want to quickly switch between worktrees and have the associated agent chat all right there. And I want all of this in one window.
I'm sure this is all possible on the command line if you spend enough time configuring tmux and vim, but I'm worried that my workflow is going to change in another few months and I'd have to do it all again. So for now I'm reinstalling Cursor 2 and I'll check back in on 3 in a few weeks.

2026-04-23 23:54:29 CEST

- reply

I saw you mentioned rootstock but I don't know what that is. Do you have a link?

2026-04-23 01:23:28 CEST

- reply

You might be interested in https://github.com/coracle-social/pomade
It's not perfect, but it does allow folks to somewhat credibly exit a custodial system with their key when they care enough to.

2026-04-22 16:15:28 CEST

I have never heard the phrase "belt-and-suspenders" before last week but Opus 4.7 finds an opportunity to use it multiple times per day.

2026-04-21 15:52:56 CEST

Another AI pattern I'm really digging lately is managing my home server with Cursor + ansible. I run a few dozen docker containers and I've always managed the server with SSH, vim, and docker CLI. I don't want an AI agent mucking around on the machine and uploading who-knows-what as context to foreign servers.
But for recent containers I have started a repo locally on my Mac where I have Claude or Composer write ansible scripts to deploy compose files and start the services. This feels like the best of both worlds to me: AI can blast out changes much faster than I can, but it doesn't have any access to the actual server and I can easily see exactly what it's going to do before I execute the playbook myself. This has allowed me to layer on additional functionality like creating a zfs dataset and ACL for every container which was too much work to do manually.

2026-04-21 15:25:58 CEST

- reply

Yes give us an example of the full stack 🤲🏻

2026-04-20 23:59:40 CEST

I spent some time over the weekend setting up a hermes AI agent on my old gaming PC. It took a lot of fiddling but I finally have some models running locally on it that make it feel like a slower slightly stupider version of claude. It feels so good to finally have a fully local stack.
There are a lot of boring chores in my life that I want AI to do and that it's probably capable of, but up until now I have refused to share much personal information with any of the big companies. I'm hoping that I can build up trust with a local agent and gradually make it more useful over time. I gave it an email address and already have it submitting some receipts for reimbursement (after approval by me) which feels like a good start.

2026-04-20 20:38:23 CEST

Instead of waiting for an app developer to fix the bug I reported I just one-shotted a replacement app with Opus. Achievement unlocked?

2026-04-16 18:21:19 CEST

- reply

What model(s) are you using with it?

2026-04-16 14:26:45 CEST

- reply

Oh I just installed hermes last night. I'm going to get it running with local qwen3. How is the browser use tool? I want to automate some boring busywork.

2026-04-15 18:40:02 CEST

Every couple months I do a race where I have some agents go off and build a feature or fix a bug while I do it myself in Cursor. The time I spend reviewing and fixing the agent's work always end up being longer and more painful, which is I haven't switched over to an "agent command-center" style of software dev.
I do kick off worktree agents here and there throughout the day to make minor changes that come up while I'm working on a larger branch. But those are side quests while I work on the main thing.

2026-04-15 18:33:58 CEST

Cursor's Composer 2 model is performing much worse for me than Composer 1 :( I feel like Composer 1 really hit a sweet spot for me between speed and quality.
For me the bottlenecks for coding with AI are:
- understanding all the code that the model wrote
- testing changes
Composer 1 really helped with the first because it could blast out small amounts of code that I could quickly review without my brain getting bored and context switching to something else. I feel like I'm an outlier in that I'm trying to stay heavily involved in the dev flow rather than having a multiple agents work on long tasks and then coming back in cold to review their work. Is anyone else using smaller quicker models in this way?

2026-04-15 14:49:02 CEST

- reply

Interesting idea to make the new people pay. I like it because readers are never going to pay much for curation.
High signal is so subjective though.

2026-04-15 04:34:53 CEST

- reply

What kinds of worklows have you built? I don't understand the bottleneck this is trying to solve for.

2026-04-13 23:38:08 CEST

- reply

Congrats!

2026-04-13 22:22:59 CEST

- reply

Cc @nprofile…pqyq @nprofile…dgdk

2026-04-10 20:32:04 CEST

- reply

Haha I wasn't expecting to see Paul Frazee in there with the deep cut.

2026-04-10 20:26:25 CEST

- reply

It's a bit out of date rn, but the broad strokes are there: https://github.com/mplorentz/horcrux/blob/main/NIP-draft.md

2026-04-10 17:54:47 CEST

- reply

You can append a NIP-05 address to a njump URL for something more readable, like this: https://njump.me/mattlorentz.com

2026-04-10 16:47:38 CEST

Do we have a NIP/tag that says to a relay "only serve this event to the authenticated author or p-tagged recipient?"
This behavior is mentioned in NIP-17 and NIP-9a and probably makes sense in a lot of cases, and I want it for my Shamir's Secret Sharing NIP.

2026-04-02 22:56:02 CEST

- reply

WebRTC tries to make p2p connections but falls back to a forwarding server if necessary.

2026-04-02 22:54:40 CEST

- reply

That's not handled at the application level. If you connect to flotilla over Tor or I2P or whatever then your IP should be obfuscated to the other participants and forwarding server as well.