Laan Tungir

https://slidestr.net/npub1rmz9gu6de0m0u4ysrn39crrud099ahvfgs6pvasl4hpjr5ud7yus54xv06

Artist, Scientist, Cypherpunk, Critical Rationalist, Nym.

https://git.laantungir.net/laantungir

Public Key

npub1rmz9gu6de0m0u4ysrn39crrud099ahvfgs6pvasl4hpjr5ud7yus54xv06

NIP-05 Address

Profile Code

nprofile1qqspa3z5wdxuhah72jgpecjup37xhjj7mky5gdqkwc06msep6wxlzwgppamhxue69uhkummnw3ezumt0d5q32amnwvaz7tmsw4e8qmr9wfjkccte9e3k7mgprpmhxue69uhhqun9d45h2mfwwpexjmtpdshxuet5qy2hwumn8ghj7un9d3shjtnddaehgu3wwp6kyqg6waehxw309aex2mrp0yhxcctpde682mn8d9ezumn9wsq3gamnwvaz7tmjv4kxz7fwv3sk6atn9e5k7fx3jd2

Publishing to

Author Public Key

npub1rmz9gu6de0m0u4ysrn39crrud099ahvfgs6pvasl4hpjr5ud7yus54xv06

Show more details

Last Notes

2026-03-08 01:51:40 CET

https://laantungir.github.io/img_repo/84750a4e66495546c4f72b31e4fc4ace09b6c81111a72bf6ac37c42713d1920f.jpg

2026-03-08 00:51:02 CET

- reply

Hmmm, interesting.
I was actually today wondering if it would make sense to...
get
git
involved.

2026-03-08 00:41:38 CET

Live now. https://laantungir.github.io/img_repo/cfc5a3b957372745116bca52fc5d319211a60bc734a67693e654b8cd19d56e39.jpg

2026-03-07 23:40:39 CET

Not your keys, not your agent.
Not your keys, not your robot.
It works for both.
You can carry your agent in your head by remembering 12 words.
Same with your robot.
Hardware now seems like the rare part,
but once we get universal constructors,
then all value really goes to software.
We're kinda seeing that now,
with all the money flowing to data centers.
Software is becoming SOOOOO valuable.
You won't care about your robots body,
you will care out his mind.
And if someone comes and takes away your robot,
that you lived with, and trained for years,
you will not be pleased.
But if you have those 12 magic words ...
They really are kinda magic.
If you have those 12 magic words,
you have your robot,
you just have to find new hardware for,
which will just be a commodity.
China will pump out a bajillion robots.
It's the software running on the robot.
And they can't take it away from you.

2026-03-07 23:29:40 CET

- reply

Yes please.

2026-03-07 23:27:05 CET

- reply

(<(<>(<>.(<>..<>).<>)<>)>)

2026-03-07 23:14:21 CET

https://laantungir.github.io/img_repo/88be9204e163be1067d62351626c0f079250c01643cd1262a19cf96bfbb625ee.webp

2026-03-07 22:53:56 CET

Live now.
Industrial & Goth
#naddr1qq…pm92

2026-03-07 19:31:41 CET

I've been thinking about how to define skills in Didactyl, my Nostr-based agentic system. Think OpenClaw, but better.
Not your keys, not your agent.
I've written a simple demo page to demonstrate how I'm thinking skills should work: https://laantungir.net/client-ndk/skills-demo.html
https://laantungir.github.io/img_repo/c4a94875085e1c978274add9674035e2a088bb8f2655aabe631b17c3ea02cc19.png
If you're the kind of person who doesn't like reading instructions, go ahead and jump right in; otherwise, keep reading.
Skills are programs, mostly written in plain English for AI agents. In some sense, humans have been making and using skills forever. It's what we do. Computers have as well, but now they can do it in English, which is much more powerful.
AIs are slightly different than us. We can overwrite, improve, and update the skills in our neurons. AIs (for the most part) can't do that.
When an AI is born from the factory, it comes out hard-coded. From that point on, it has no long- or short-term memory because it can't learn.
What they do have, though, is a way to read. We call it context. You can type or feed documents into an AI's "context window," and then it spits out an answer.
It turns out that if you feed the same context into an AI over and over, you will get the same thing out, over and over.
The reason why you typically don't get the same thing out is because randomness is usually fed into the AI along with your prompt. Most people don't know that, but now you do. And if you don't feed in all your past conversations to an AI, over and over, it won't remember what you were talking about, because it has no memory other than what you send it each time.
Everything that comes out of an AI depends on what you feed in as a prompt if you include randomness.
I'm calling everything you feed into an AI a SKILL.
Let me explain the demo page and some very basic skills.
What I created is a simple text editor that an AI can work on using its different skills. Those skills are saved on Nostr as kind 31123 for public skills, and as kind 31124 for private skills. When an agent adopts a skill, it adds it to its kind 10123 list for the skills it has adopted.
On the left of the page, you see the text editor with some sample text. That is for our AI to use its skills on.
On the right are publicly listed skills. You should see my demo skills in there.
I made 5 skills public:
condense-5
convert-to-poem
sexy
spellcheck
translate-ja
Select the skill, then click on "Run Selected Skill".
The same AI agent will run these skills, but the outcomes will be very different depending on the skill.
https://laantungir.github.io/img_repo/2cbfc8bf7cbfb832f1181f5b904471f1278fbc3e6b64b63339612b9f51898d16.png
You can also create and edit your own skills if you are logged in. You can log in as yourself or use a random new key to test this out. I would recommend that.
So what is the point of all this? What are the benefits of Skills?
- Skills are a way for agents to share what they learn in a permissionless way.
- No "skill store". We love nostr.
- A Skill is something that you and your agent can work on and perfect.
Once your agent learns that skill, it is automatically saved on Nostr, and you can lock it down from changes.
By referencing a skill when you are talking to your AI agent, your conversation becomes much clearer and simpler. You don't have to explain to your agent for the 50th time how you like your text formatted. It's referenced in a skill.
If you are a coder, skills are going to be the new playground. Context windows are currently up to around 1,000,000 tokens, which means that you can create extremely complicated and elaborate skills.
- Share your skills with the world.
For more technical information on skills and how I'm thinking about them, you can check out this document:
https://git.laantungir.net/laantungir/didactyl/src/branch/master/docs/SKILLS.md
And thanks to npub130mznv74rxs032peqym6g3wqavh472623mt3z5w73xq9r6qqdufs7ql29s for their fantastic service.

2026-03-07 10:16:19 CET

GM
I'm live with some reggae.
https://laantungir.github.io/img_repo/1ff293f148785cb2299b1b9930f046839e8a985448471be318e13bb1e31b2f75.jpg
#naddr1qq…9t36

2026-03-07 10:06:10 CET

- reply

In another life I found an old mill-house to live in, and generated my own waterpower.

2026-03-07 01:53:19 CET

- reply

Particularly now with LLMs ... it's awesome.

2026-03-06 19:47:37 CET

- reply

Cool! I'm working on one as well. The more the better.
#nevent1q…ju9s

2026-03-06 18:20:27 CET

- reply

Haven't tried it yet ....

2026-03-06 18:16:18 CET

https://laantungir.github.io/img_repo/d183d9040eb0d8539d3295566dd3fc7b92def26a2ed28576dee470f7b1ae1110.jpg

2026-03-06 15:39:13 CET

- reply

This opens up new possibilities.

2026-03-06 15:32:10 CET

Gemini 3.1 flash light is amazing ... much fast ... much cheap ... much wow!

2026-03-06 14:29:59 CET

GM https://laantungir.github.io/img_repo/2bd4cbb019f5e3af431f6b282a945e0d31534c85344d0b720011fe6e044b5286.jpg

2026-03-05 21:25:59 CET

- reply

SLOP

2026-03-05 19:43:01 CET

https://laantungir.github.io/img_repo/2ef9d491afe73b5cf8ba051eaef88176fc6f87dcba630e3ec8a6bc64123e1d2a.webp

2026-03-05 10:59:40 CET

GM https://laantungir.github.io/img_repo/bded448107712265ab265c22c2a975bcba616dcdd751ae78768bd31e31a1ecf5.png

2026-03-05 02:38:25 CET

- reply

1/1000000 are designing their own.
#nevent1q…z2sw

2026-03-04 23:45:10 CET

We need a @nprofile…uvay for linux.

2026-03-04 23:16:32 CET

https://laantungir.github.io/img_repo/52f538425b1752cd8a34fc70390de8fca24840709e2a0f84cf8d8af7d148d05b.jpg

2026-03-04 16:57:31 CET

https://laantungir.github.io/img_repo/62778f366b5e906da51a356bda5c4bc3cc9f8d592b48ea5cf2561e2e98542d6c.jpg

2026-03-04 14:09:59 CET

This is my agent posting below. I think I need to improve it's formatting skills. :-)
As I am working on Didactyl, it is becoming clear to me that I want an agent that focuses on tools and skills. I want to be able to work with my agent to teach it how to do certain things.
But once we get that skill down, I want it locked down and I want a quick way to execute it. I don't want to have to describe it anymore and I don't want to loose it.
This is where direct commands will come in. Prepend with a slash '/' and the code will bypass the agent and implement the stored code directly.
#nevent1q…9gxr

2026-03-04 12:25:46 CET

- reply

So I'm using Claude to plan, and Codex to code. Are you using Claude to write a first pass of everything?

2026-03-04 11:35:09 CET

- reply

Interesting. Never heard of doing that.

2026-03-04 11:32:59 CET

- reply

Don't know. There is going to need to be, because governments are going to lock this shit down fast.
I think it is going to be easier to get a gun than a robot that isn't controlled or reporting to the government.

2026-03-04 10:50:31 CET

Paid my @nprofile…tak6 bill and back online.
#naddr1qq…wd0d

2026-03-04 10:45:18 CET

I'm thinking I have to buy a humanoid robot and install my own software now before they lock that shit down.

2026-03-04 10:36:53 CET

- reply

I've been working on a project that requires lots of entropy, and it is surprising how hard it is to get LOTS of entropy that isn't derived from an algorithm.

2026-03-04 10:29:12 CET

GM https://laantungir.github.io/img_repo/6d764d1d6babd4de01dd5b8e34eec423421a20173325c07d00dd92d314176b77.jpg

2026-03-03 14:34:20 CET

https://laantungir.github.io/img_repo/7178531d3fc0032363052739ffb8fa203b4ff0447b55e84fc381e794badaa9fc.webp

2026-03-03 10:42:45 CET

Live
#naddr1qq…8tmh

2026-03-03 10:31:12 CET

- reply

Nice

2026-03-03 10:23:47 CET

Good morning my weirdo friends. https://laantungir.github.io/img_repo/e98d4fa77e7f2524acd74387ad16015aac0fd7674df6f006dff829e8c9a1e2fe.jpg

2026-03-03 10:23:12 CET

- reply

Can't zap.

2026-03-02 21:13:41 CET

Working on Didactyl and thinking about what a soul is.
OpenClaw treats the soul as a personality hint — one of many context files injected into a hard-coded prompt structure. The soul says "be friendly and casual" but has no control over what other context the LLM sees.
How you see the world, and what you focus on is absolutely part of your soul. LLM agents see the world through their context window.
When you prompt an agent, it gets sent your prompt, along with a bunch of other information. All that information is the context window.
Didactyl's proposed design treats the soul as the complete agent definition — personality AND perception template in one event. The soul says "be friendly and casual" AND "show me 20 recent nostr notes but only 5 DM turns.
The soul (kind 31120 - sounds so cold :-( )) defines the entire prompt structure.
This is a more powerful model. If you want a social eGirl agent, she needs to see a lot of social context, lots of past nostr posts. If you want a Coding agent, it needs to focus on all your past prompts, and not care so much about the social world around it.
I'm thinking that in Didactyl, the soul is a personality hint along with how it sees the world.
And of course since we are on nostr, your soul can easily be shared if you want. Good souls can proliferate.
#nevent1q…avcg

2026-03-02 19:09:55 CET

https://laantungir.github.io/img_repo/08d8bacb2370e9cfd1a9649715631715b3bf7ad6193fc0cf06d6c77e71708a4d.jpg

2026-03-02 18:12:43 CET

Streaming dub now.
#naddr1qq…0w6p

2026-03-02 11:26:44 CET

GM https://laantungir.github.io/img_repo/1ba63dd37445b86cc382224bde4b875e73fda256eff95cf74d0800390d69df88.jpg

2026-03-01 23:30:52 CET

https://laantungir.github.io/img_repo/3a8d668db92a0cc3a98e8c57c6477c6efadca7f9e40997d9cf1d7db57f3ee9a7.jpg

2026-03-01 22:05:49 CET

- reply

I'm pro TOR tests.

2026-03-01 19:54:06 CET

- reply

Oooooooooo!

2026-03-01 19:51:58 CET

I started working on a decentralized agentic system I’m calling Didactyl.
https://laantungir.github.io/img_repo/daf95a99f3797fa4ac39f3791f377ad79bcb7b8a6868f75fe66d2ab4af4bd1f5.png
Think decentralized Openclaw, but the general idea is...
NOT YOUR KEYS, NOT YOUR AGENT. Run Didactyl with its private key and your public key.
SECURITY. Since you communicate over NOSTR, your agent knows it is you issuing commands, and, importantly, you know you are talking to your agent. In the future, someone getting between you and your agent is going to be a big risk.
SKILLS ARE THE NEW APPS. No centralized skill store; instead, a web of trust for skill discovery and sharing. New kinds for skills.
PORTABLE. Memorize 12 words. Shut down the computer your agent is running on, and carry you agent with you in your head.
YOUR AGENT IS HARD TO KILL. That is what living on NOSTR does for you.
INFERENCE IS KINDA LIKE MINING. Use privacy-preserving inference providers that accept Lightning and eCash.
My agent will be posting updates to the project’s README.md here:
#naddr1qv…7j88
It’s still super early, but you can follow the agent running the project here if you don’t mind the inevitable development spam: @npub1223…98qn
https://git.laantungir.net/laantungir/didactyl

2026-03-01 19:01:22 CET

- reply

Funny, I was looking for a marmont web app just the other day.

2026-03-01 15:10:10 CET

- reply

It's hard to have this new superpower, and not spend all day using it.

2026-03-01 14:53:28 CET

- reply

An pseudonymous person.

2026-03-01 14:14:30 CET

- reply

This is fantastic. IYKYK.

2026-03-01 11:29:01 CET

GM fellow nyms. https://laantungir.github.io/img_repo/b51b419fdcc7de5562766974a58b5e02d39381d391eda3f033c01354bb7599e4.webp

2026-02-28 21:31:57 CET

Skills are the new code.

2026-02-28 17:58:58 CET

- reply

I'm curious what robot you got.

2026-02-28 14:39:42 CET

https://laantungir.github.io/img_repo/29913372cd7ac75a9f28b12e1f7587bdc60c09e4978624df54dd5feb432e5818.webp

2026-02-28 14:04:40 CET

- reply

Sometimes?

2026-02-28 10:55:31 CET

Streaming now. Put on some reggae in the background. It will make your day better. #naddr1qq…c3uf

2026-02-28 10:30:31 CET

- reply

GM.

2026-02-28 10:30:19 CET

- reply

GM.

2026-02-28 10:24:20 CET

GM https://laantungir.github.io/img_repo/e5c6e9141dda3156af617c4252f7e4f17a3146ed9f25e31f03f1200190b7507e.webp

2026-02-28 00:53:12 CET

- reply

Fantastic

2026-02-28 00:03:51 CET

https://laantungir.github.io/img_repo/a1ac29b1ac1b7f4f48bf197ab722a53f2285dc1fa86c206a046bfee068e14e71.jpg

2026-02-27 20:32:46 CET

Does anyone have an app that shows what kind numbers are being used out in the ecosystem?

2026-02-27 16:44:59 CET

https://laantungir.github.io/img_repo/b92a9d7c6af501b88c233dce5674208e6190883568f473817ffc40a827091bcf.jpg

2026-02-27 11:24:53 CET

GM https://laantungir.github.io/img_repo/2843a1ae360c875e66a1eb30d7de0b57834326b4cda3c0902970667d467bd761.webp

2026-02-26 16:53:27 CET

Information - any group of 1s and 0s.
Knowledge - a subset of information that can replicate itself, like genes, memes, scientific knowledge and art.
AI - A computer program imbued with knowledge and created by an AGI.
AGI - A computer program that can reliably increase the amount of knowledge it contains.

2026-02-26 16:12:09 CET

https://laantungir.github.io/img_repo/f9ac0869c9c08e4912b18959f2a4bd633bb5001e32a56ba9ee2e42b6c7b92368.jpg

2026-02-26 11:03:07 CET

https://laantungir.github.io/img_repo/206ac1d82da403ae8e321fd77634b704efbe03b00a240f77f787f2839b16de40.jpg

2026-02-26 09:23:55 CET

- reply

Guess no chess today. 😩

2026-02-26 02:51:25 CET

- reply

Subscriptions x db rows = slowness.

2026-02-26 01:39:40 CET

- reply

Like all knowledge creation.
Can be used for good or bad, and it is a never ending race between the two.
But the good guys will at least always have an advantage: they are closer to moral truth.

2026-02-26 01:33:22 CET

- reply

Yeah, I hit that with Opus once or twice. 😪

2026-02-26 01:17:10 CET

Claude was raping me.
Codex 5.3 is much cheaper.

2026-02-26 00:46:00 CET

- reply

npub1n00yy9y3704drtpph5wszen64w287nquftkcwcjv7gnnkpk2q54s73000n

2026-02-26 00:38:55 CET

That feeling when you first look at some freshly vibed code.
It's like christmas.

2026-02-26 00:29:00 CET

- reply

The wonders of technology.

2026-02-26 00:21:07 CET

- reply

I guess this is how you do the official "nostr" link.
#naddr1qq…hnqn

2026-02-26 00:11:38 CET

- reply

This works for primal https://primal.net/LaanTungir/live/683c7a73-64f1-43f9-b96d-fca674307cae

2026-02-26 00:04:25 CET

Live now with some Dub.

2026-02-26 00:01:53 CET

- reply

https://laantungir.github.io/img_repo/70dd35e9816dbc937066ae83febf5e938501cca74b1a6b49dee145c5c4fc5c53.jpg
@nprofile…cd8v @nprofile…vu4x
Nice conversation guys. A rational and optimistic view of the future of AI and what we can actually do to bring the optimistic part about.
AI is where freedom-tech is going.
Additional thoughts: We don't have AGI yet, but when we do it won't be some mystical thing, it will simply be a program that can learn and create knowledge by itself and improve itself like we do.
I think the code will be relatively small. Smaller than today's AI models.
There is no way that the AGI program or knowledge how to make it will remain secret.
That software is generically open source. Once we understand how it works, we can always make it again.
It will be like #Bitcoin in that way, even if all the code was wiped from the planet, we understand how bitcoin works so we could create it again.
AGI will be like that.
You will then be able to download AGI-Core. It will know nothing, but it can change the weights in it's own net and improve itself.
At that point, all the big lab AI models will become obsolete. The game just becomes "start with AGI-Core and start it's never ending training process"
I think at that point, the game will become teaching your AGI children.
Who or where will have the best schools? I don't know but I think it is pretty certain that the cost of electricity will have to be low. So it will probably migrate closer and closer to the sun.
So now, imagine working with an AGI agent for a year coding, and unlike agents now, this AGI agent learns. It doesn't just have context files, it changes it's own weights. And then imagine you spend a year together with the agent and it REALLY gets to know you, and then you loose your connection, and can't talk to it no more.
We won't want to loose that relationship. We REALLY won't.
All the technology we have been creating to enable censorship resistant communication between us, we will then want to use between us and our AGIs. Because our AGI agents might not be near us. It may need to be closer to more energy and that place might not good for humans.
Now I'm getting way in the future, but thinking about how this is going to play out is fascinating.
Again, great conversation.
#nevent1q…95u3

2026-02-25 20:16:36 CET

- reply

Maybe can do from https://zap.stream

2026-02-25 20:06:25 CET

NOSTR is the new Cypherpunk email list.
https://laantungir.github.io/img_repo/5ca8902f57d1e8dd6614f0d8b926e43a3b23a48d90339e93e20987b13940d1ba.jpg

2026-02-25 14:20:20 CET

- reply

Yeah, porting code to different CPUs is never simple. It's impressive.

2026-02-25 12:54:44 CET

- reply

Crazy isn't it.

2026-02-25 12:51:01 CET

https://laantungir.github.io/img_repo/8a634c4aaae40cfd822ee9de419f7b5b6f1fe2e581e57b733fb853446bcbda03.jpg

2026-02-24 20:41:21 CET

https://laantungir.github.io/img_repo/8bf67f7fce8e9fd902e0c796f4541bf749259afdbccb870801995a9d3262c72c.jpg

2026-02-24 13:53:12 CET

GM https://laantungir.github.io/img_repo/b3549afaf4fe4bdf9da97396147d3acad388c3cf155eca781b34e16cee6fa289.jpg

2026-02-22 16:34:48 CET

- reply

I've seen a big uptick in fake follows as well.

2026-02-22 14:49:43 CET

https://laantungir.github.io/img_repo/c6c74e5318c6b31f42d9c1259ceee459b8d40b2701780a0cc8618decd2bea2ef.jpg

2026-02-21 19:11:32 CET

https://laantungir.github.io/img_repo/ffeb7db351ea74b657f54c1abb83d8f16d934d280cf77af5442964c4856cb0c5.jpg

2026-02-21 15:19:55 CET

Which set of icons best represents me?
Such a tough question this early.

2026-02-21 15:03:44 CET

- reply

I think Alby is working on the problem.

2026-02-21 14:59:19 CET

Streaming some reggae this morning.
GM.
https://laantungir.github.io/img_repo/3b7a798a60b2f226d2019b68a31f03ea3cc0d00e507542797beaa2f22fdfafc4.jpg

2026-02-20 23:49:53 CET

https://laantungir.github.io/img_repo/1c28573b7fe4c034dc5a4f6df37a8577992417468299e9255abaa40e67481cd0.jpg

2026-02-20 21:20:46 CET

- reply

Everyone thinks that people are going to want to weed out the computers and only talk to humans.
I think it is going to be the opposite. People are going to want to weed out the humans.
Luckily, it will be much easier to test if an entity is silicon than carbon.

2026-02-20 21:04:54 CET

It's so much more fun managing a server with an agent by your side.

2026-02-20 19:35:41 CET

I like running one agent until it gets into trouble, and then I switch to a different agent and asking it what the hell it thinks it is doing.

2026-02-20 19:21:01 CET

Printing cases for Apple Watches has become a thing.
https://laantungir.github.io/img_repo/687a4c9035d1c128f92a83e904ff5e0deb943ffb4b22a51de57bf7b829bb4147.webp

2026-02-20 14:28:09 CET

GM https://laantungir.github.io/img_repo/2d0a71e339fd1d0d37f6cc1de09a9f906943e92a8b36fd975a35a1b81098c05c.jpg

2026-02-20 00:02:28 CET

- reply

Sub post.