iiinova — where your digital self keeps evolving

Philipp Pohl-Krüger @matheus23.com · 29/09/2026

There are some interesting ideas for future jobs by modern monetary theory proponents: The Job Guarantee. The idea is to have the government fund community or care work at livable wages.

010

Reposted by Philipp Pohl-Krüger

rklaehn.bsky.social @rklaehn.bsky.social · 29/09/2026

We have released a new version of @iroh.computer. Nothing major in terms of features, but we did manage to significantly reduce the binary size. So now my entire embedded device zoo is running on iroh from crates.io. No more patch needed. www.iroh.computer/blog/iroh-1-...

01059

Philipp Pohl-Krüger @matheus23.com · 28/09/2026

Relatable. For real though, do you think it might get activated because of the word "execute"?

120

Reposted by Philipp Pohl-Krüger

Evangeline Rome 🌵 🦎 🌱 🐢 🐍 🐸 @evadeva.bsky.social · 24/09/2026

@iroh.computer Was "Dial Keys, not IPs" inspired by reticulum.network? And what would be the differences between them?

Reticulum does not use source addresses. No packets transmitted include information about the address, place, machine or person they originated from.
There is no central control over the address space in Reticulum. Anyone can allocate as many addresses as they need, when they need them.
Reticulum ensures end-to-end connectivity. Newly generated addresses become globally reachable in a matter of seconds to a few minutes.
Addresses are self-sovereign and portable. Once an address has been created, it can be moved physically to another place in the network, and continue to be reachable.
All communication is secured with strong, modern encryption by default.
All encryption keys are ephemeral, and communication offers forward secrecy by default.
It is not possible to establish unencrypted links in Reticulum networks.
It is not possible to send unencrypted packets to any destinations in the network.
Destinations receiving unencrypted packets will drop them as invalid.

121

Philipp Pohl-Krüger @matheus23.com · 27/09/2026

We just ended up realizing that dialing by key alone is already a super useful abstraction. Maybe more things should be built like this. I'm glad to see that's how reticulum is designed as well. Not sure about exact differences. Someone would have to educate me on reticulum.

120

Philipp Pohl-Krüger @matheus23.com · 27/09/2026

It wasn't! We aren't actually super familiar with reticulum and we weren't when we named it. The concept mostly came out of a need for stable addressing in a world where neither IP nor DNS was possible to address a device adequately and securely.

150

Reposted by Philipp Pohl-Krüger

Isaac Corbrey @isaaccorbrey.com · 24/09/2026

I swear @iroh.computer is erasing basically every network annoyance I could have possibly encountered when ideating about decentralized systems

0474

Philipp Pohl-Krüger @matheus23.com · 22/09/2026

No. But you only need to give it one more piece of information: the relay URL (or the DNS pkarr server to resolve it, but at that point the relay URL is likely easier)

100

Philipp Pohl-Krüger @matheus23.com · 22/09/2026

Relays don't have "a table" and don't use any pkarr mechanisms which is why I can't parse your last question.

100

Philipp Pohl-Krüger @matheus23.com · 22/09/2026

First, relays are oblivious about DNS resolution. Second, the pkarr mechanism is mainly used to resolve the "home relay" of an endpoint. You can contact as many DNS pkarr servers as you want in order to try resolving a home relay. Knowing the home relay is mostly what we need for connectivity.

110

Philipp Pohl-Krüger @matheus23.com · 22/09/2026

You don't need agreement on the relay set. You could add another record to the above that tells you about the pkarr DNS resolver used. That alone would make it universal. Endpoints on different relays can talk to each other. There are other ways of slicing the problem, but that'd be one of them.

130

Philipp Pohl-Krüger @matheus23.com · 19/09/2026

I saw that. I'm a bit slow to respond at the moment - was on vacation this week. I kept that project private mostly because it was heavily LLM assisted. If you want to build in public that's great :) if you decline the invite I respect that :)

010

Philipp Pohl-Krüger @matheus23.com · 11/09/2026

Excited to talk about iroh and networking :) We'll try to record this! ✌️

0210

Reposted by Philipp Pohl-Krüger

𝓐𝓾𝓫𝓻𝓮𝔂 @aub.bsky.social · 09/09/2026

"I do well with explicit instruction and well-defined tasks with unambiguous success criteria" well, agents do too & they're faster than you, so unfortunately all we have left is [looks at clipboard] vague direction towards poorly-defined ends without any clear feedback loop for success or failure.

2929

Philipp Pohl-Krüger @matheus23.com · 08/09/2026

Just looking up some literature. This is clearly discussed. Mostly as a way of allowing a model to spend more compute on harder problems, not on the read vs. write axis, which I guess is a more correct framing.

001

Philipp Pohl-Krüger @matheus23.com · 08/09/2026

I wonder if chain of thought was so successful because it makes token generation use more computing resources than prompt processing. Reading a source file and producing a source file are otherwise forced to use the same amount of compute per token.

130

Philipp Pohl-Krüger @matheus23.com · 07/09/2026

Yeah I'm doing the first talk :)

110

Reposted by Philipp Pohl-Krüger

mr. TIM @timkellogg.me · 07/09/2026

it is sooooo cool that we have an actual explainable algorithm here on bsky

116327

Reposted by Philipp Pohl-Krüger

Niklas Korz @niklaskorz.de · 06/09/2026

Official announcement will follow next week: we're meeting on Tuesday, October 7, for our next @NixRust meetup in #Mannheim. We still have time for a second talk, so if you have a topic or project related to #RustLang or #NixOS that you'd like to present, get in touch!

223

Philipp Pohl-Krüger @matheus23.com · 04/09/2026

It's possible I'm over indexing on what the current thing is good at...

120

Philipp Pohl-Krüger @matheus23.com · 04/09/2026

Maybe 4 parallel any time max at the moment. Generally more fire and forget stuff. "Give us a morning briefing" and similar things.

120

Philipp Pohl-Krüger @matheus23.com · 04/09/2026

Id like it to do better than a 256GB Mac Studio M3 Ultra that we already have... which seems to be difficult 😬 Those two 3090s wouldn't even get close in terms of models we could run, given it's just 24GB VRAM.

120

Philipp Pohl-Krüger @matheus23.com · 04/09/2026

If you had $8k USD to build a local LLM server, what would you spec? Seems pretty hard to me. Maybe one of you has ideas?

330

Philipp Pohl-Krüger @matheus23.com · 02/09/2026

Seconding this. I'm feeling a lot safer since I've had to get set up with contact lenses for diving.

000

Philipp Pohl-Krüger @matheus23.com · 27/08/2026

Billions of floating point operations for these things to do this will never stop being funny to me

$letta code terminal excerpt: • Run Verify Qwen reasoning control and parser behavior on Vulkan └ { "content": "1517", "reasoning": "37 × 41 = 37 × 40 + 37 × 1 = 1480 + 37 = 1517\n",$

0151

Philipp Pohl-Krüger @matheus23.com · 26/08/2026

@nonbinary.computer woah the jacquard repo pull is 432MB at the moment. Seems like that's mostly the binaries subdirectory history. Any idea if that could be improved?

140

Philipp Pohl-Krüger @matheus23.com · 17/08/2026

What do you use it for/like about it?

110

Philipp Pohl-Krüger @matheus23.com · 17/08/2026

If you're using itoh-gossip, then this builds out a net over the peers in the same topic. Not everyone is connected to everyone else, but you keep ~5 connections to others in the same topic alive, and have a list of backup peers to contact if some of the active ones fall out.

150

Philipp Pohl-Krüger @matheus23.com · 15/08/2026

It gave two answers in its response. One was the over engineered circle, the other was a snippet to copy a plain circle as SVG.

010

Philipp Pohl-Krüger @matheus23.com · 15/08/2026

Ah yeah. Only Germany and Austria it turns out.

010

Philipp Pohl-Krüger @matheus23.com · 15/08/2026

I did buy DRM free from thalia.de once (and it's the only time I did buy there online)

100

Philipp Pohl-Krüger @matheus23.com · 11/08/2026

Gooooood thread feat. cryptography terminology, Mark Miller, Granovetter Diagrams and a little bit of iroh.

061

Philipp Pohl-Krüger @matheus23.com · 10/08/2026

I don't remember if this made it in yet, but I don't think we implemented proper round-robin for different paths in noq yet. Making noq aware of multiple transports/sockets would also be helpful for these endeavors I think.

000

Philipp Pohl-Krüger @matheus23.com · 08/08/2026

Debuggers are overrated :) (The tool, not the informal job description lol)

080

Reposted by Philipp Pohl-Krüger

zooko2.bsky.social @zooko2.bsky.social · 07/08/2026

Dear American tech leaders: There is no way to pause or slow down offensive cyber research without hobbling (American) cyber defenders. But there *are* ways to accelerate cyber-defense without accelerating cyber-offense! ⤵️

172

Philipp Pohl-Krüger @matheus23.com · 06/08/2026

I still need to get used to random iroh mentions in my timeline.

120

Philipp Pohl-Krüger @matheus23.com · 06/08/2026

Oooh and this is how you — - – —

000

Philipp Pohl-Krüger @matheus23.com · 06/08/2026

I just figured out how to type – on my keyboard. This is the wrong time to adopt using – in favor of -, isn't it?

320

Reposted by Philipp Pohl-Krüger

Brooklyn Zelenka @expede.wtf · 04/08/2026

Excited to have a ✨ guest post ✨ from @djwyen.bsky.social in our @inkandswitch.com Keyhive lab notes! The local-first setting is new to a lot of cryptographers; here Derek shares some of his insights working on the BeeKEM paper, cross-fork security (CFS), and fundamental protocol design tradeoffs 🐝

inkandswitch.com

06 · E2EE in the Local-First Setting

Examining design considerations for developing cryptography for the local-first setting.

13710

Philipp Pohl-Krüger @matheus23.com · 04/08/2026

Aaaaaaah We need a UCAN system so that sending emails can finally be commoditized in a standardized way and I can give permission to execute my automations to others in a local-first way.

070

Philipp Pohl-Krüger @matheus23.com · 04/08/2026

Er erwähnt das Fediverse nicht, aber: > Über das Eurosky-Netz sind Identitäten, soziale Beziehungen und Inhalte nicht mehr an eine einzelne Anwendung gebunden. Soweit ich das verstehe, ist das also etwas anderes, was im Fediverse nicht geht.

010

Philipp Pohl-Krüger @matheus23.com · 03/08/2026

Yeah - it exists. You can download from multiple providers: docs.rs/iroh-blobs/l... (any iterator of EndpointIds implements ContentDiscovery)

docs.rs

Downloader in iroh_blobs::api::downloader - Rust

API documentation for the Rust `Downloader` struct in crate `iroh_blobs`.

120

Philipp Pohl-Krüger @matheus23.com · 03/08/2026

Hey, @niklaskorz.eu, unfortunately your Bluesky DMs are closed (there was a bug recently that set lots of accounts to no-DMs). I'd be interested though! Availability is limited in September between 12.09. and 27.09, otherwise schedule is open.

120

Philipp Pohl-Krüger @matheus23.com · 03/08/2026

Eh I think it's fine. Whoever actually ends up not abandoning the project gets to keep the name in the minds of the users lol We haven't touched that experiment for years.

010