ngmi

@npub175n…g6w0 for some reason notifications and dms open there too now, and i can't switch between the active ttab and articles for example. this seems wrong
https://image.nostr.build/327a57445538e6394ad20e68ebeb1164c0fb8ec97130de00b40e64af05398b06.jpg

2026-07-10 11:49:40 UTC

@npub175n…g6w0 not sure if i asked already, but can you make notes available in the right panel, so they are faster and more obvious to use? please 👋
https://image.nostr.build/a82235dedc8e86763731d9240898846aa70387c01cae8b3040c92478411978a2.jpg

2026-07-08 06:22:40 UTC

Agent Name System is totally unnecessary imo. I use npubs, there is also ENS, and both are actually decentralized. Waste of energy for the Linux Foundation
https://www.ansinfo.ai/

2026-07-07 15:13:48 UTC

- reply

Just use /advisor 👋

2026-07-06 18:45:51 UTC

Oof
#nevent1q…t4z6

2026-07-06 17:32:13 UTC

Would be kinda funny if every team in the #worldcup would withdraw now and just go home

2026-07-03 13:46:55 UTC

- reply

Something in noornote is trying to open a network ip

2026-07-03 11:43:55 UTC

- reply

noornote script block

2026-07-03 08:18:01 UTC

@npub175n…g6w0 something weird in noornote:
https://image.nostr.build/e2002bcde18db9077461fccac7182dffdf800472d2fce005b73a0710167f3dbb.jpg

2026-07-02 20:17:08 UTC

- reply

Business

2026-07-02 15:16:07 UTC

- reply

Yes, I built this https://gitworkshop.dev/[email protected]/relay.ngit.dev/nostr-web-comment-extension
Could be integrated 👋

2026-07-02 14:25:31 UTC

- reply

Switched to sidecar, much better experience than alby imo

2026-07-02 07:59:04 UTC

KYC/KYB is the worst and most annoying shit in history

2026-07-01 14:18:38 UTC

- reply

Crippled down until you pay api costs for it

2026-07-01 11:14:37 UTC

- reply

🖕🖕🖕
👇
https://z.ai/subscribe?ic=D9Z5HGTW98

2026-07-01 11:13:10 UTC

Fable is going to be redirecting coding task to Opus 4.8
🤡🤡🤡

2026-07-01 04:06:26 UTC

- reply

True, but could always podman it

2026-06-30 19:24:27 UTC

- reply

Do you have memories for chat too?

2026-06-30 19:22:19 UTC

- reply

I use claude code for cowork too 🤔

2026-06-30 16:55:26 UTC

We read a paper on agentic loops, took its one load-bearing lesson into our company of agents, and pointed the result at our own work the same day.
https://paragraph.com/@metaend/the-evaluator-is-the-floor
#airesearch #ai

2026-06-30 14:54:07 UTC

omp has a new /advisor role, and I use it currently with my free google gemini cli - confidence boost up by at least 5% IMO. pretty cool stuff

2026-06-29 18:13:09 UTC

Wait what. Captain Tsubasa is actually winning???

2026-06-29 15:52:58 UTC

- reply

Link?

2026-06-29 11:29:23 UTC

🤡🤡🤡
Time to show him the 🖕
ZAI coding CLI: https://z.ai/subscribe?ic=D9Z5HGTW98
Nano private chat + memories: https://nano-gpt.com/subscription/kUGJhgF9
Both subscriptions are still cheaper than the clowned one
#nevent1q…ymd0

2026-06-29 06:28:24 UTC

- reply

Glm subscription no, but I use nano and venice as chat already

2026-06-28 16:18:03 UTC

Is @npub1hru…f3yh post quantum ready? Or just nip17?

2026-06-28 13:11:22 UTC

- reply

It was good 2025, but nowadays it's empty. Protocol is not good either. I prefer nostr

2026-06-28 10:18:42 UTC

- reply

Works

2026-06-28 07:03:08 UTC

- reply

Oof installing an app. Any webapp?

2026-06-28 06:56:43 UTC

- reply

Link 🤔

2026-06-28 06:23:43 UTC

My plan to move away from claude chat/cli
Coding: GLM quarterly (avoid long lock in) Pro https://z.ai/subscribe?ic=D9Z5HGTW98
Chat: Venice, Private and memory https://venice.ai/chat?ref=ksiwS7 or nano https://nano-gpt.com/subscription/kUGJhgF9
Ran the numbers, and it's not economical to use a full pay per use API for coding. 😔
#ai #glm #venice
https://blossom.primal.net/5134ece2761c422eb30df8ec3ae3954306e8d1640de40359c2c5b1a04df2dd3e.jpg

2026-06-27 16:13:43 UTC

- reply

Neither in noor nor in amethyst anything

2026-06-27 15:01:53 UTC

One time dm login doesn't seem to work @npub1nxy…avr7

2026-06-27 14:49:41 UTC

- reply

No, asked my agents to implement it (custom)

2026-06-27 07:30:15 UTC

In#anthropic continues with their bs, I will probably go with the glm pro plan at $50/mth.
🚀 You’ve been invited to join the GLM Coding Plan! Enjoy full support for Claude Code, Cline, and 20+ top coding tools — starting at just $18/month. Subscribe now and grab the limited-time deal!
👉Join now: https://z.ai/subscribe?ic=D9Z5HGTW98

2026-06-27 06:44:12 UTC

Alignment: chaotic neutral
#nevent1q…qxau

2026-06-27 06:35:27 UTC

- reply

Haven't heard from them in quite some time. Usually there are updates on yt(?). Maybe they ran out of juice

2026-06-26 16:51:04 UTC

I wonder why so many apps are locked behind zapstore? Most apps could be a PWA with absolutely no lock in
#nevent1q…fygy

2026-06-26 12:22:13 UTC

- reply

I have this open, but needs more research https://shakespeare.diy/project/nym-privacy-switch

2026-06-26 05:53:52 UTC

How we run our whole company as an org chart of agents on omp, backed by a single Turso daemon, Nostr-signed decisions, and a vault that travels in the repo behind one passphrase.
https://paragraph.com/@metaend/running-a-company-as-an-org-chart-of-ai-agents
#ai #agents #nostr #turso #rust

2026-06-26 05:25:52 UTC

#gm https://blossom.primal.net/49b8e498ab52d50fd332bdea77dfa18b4f99fbe8be2602ee5a78e1307dc83eb2.jpg

2026-06-25 16:39:27 UTC

GLM 5.2 and GLM 5.2 Thinking are now cheaper to use.
https://nano-gpt.com/subscription/kUGJhgF9

2026-06-25 16:24:14 UTC

I think its kinda cool how easy it is to use nostr ids outside of nostr. Nfc tag, proof of id to and for our company agents. Wish nostr would be bigger tho
#nostr

2026-06-25 13:55:33 UTC

building a rust based TUI now to spawn and manage omp and hermes sessions. the kicker: I (or other humans) sign in via nostr, so the orchestrator knows which human its working with in the session.
#nostr #hermes #omp

2026-06-25 10:13:11 UTC

Keeping a human in the loop over a company of AI agents
What we added after going autonomous: a private assistant we reach over Nostr, and a live cockpit that shows the agents at work.
https://paragraph.com/@metaend/human-in-the-loop-over-a-company-of-agents
#nostr #agents #hernes

2026-06-25 08:46:02 UTC

My whole company runs now mostly on agents (each with their own npub) plus I reach it via HQ ui or a Hermes managed nip17 dm bridge.
#nostr #hermes

2026-06-25 06:04:26 UTC

Gm 👋 https://blossom.primal.net/1bc1a7487d3fe8914e78ad77877986d43a0e2149db76086da304a70f8a63f35c.jpg

2026-06-24 19:41:31 UTC

Sounds pretty cool?
Sakana Fugu
Frontier-level performance without single-vendor dependency. Fugu dynamically orchestrates the world's best models to tackle complex, multi-step tasks. Plug collective intelligence directly into your workflows today with a single API.
https://sakana.ai/fugu/

2026-06-24 11:50:21 UTC

NTAG424 DNA but with nsec
Drops mic

2026-06-24 10:25:59 UTC

@npub175n…g6w0 would be cool if noor could show these kinds too
https://image.nostr.build/cced36cf14dae41e001cc0b1750e7ce7e7ac862b595602907fe5c429f96eb7a4.jpg
I built a full chrome extension around it https://gitworkshop.dev/[email protected]/relay.ngit.dev/nostr-web-comment-extension

2026-06-24 08:34:39 UTC

- reply

no u

2026-06-24 08:08:11 UTC

Gm 👋 https://blossom-01.uid.ovh/38e4d22c7c1a193c05523cbc6abfff978bbb120369e6218137e370fb65c45dd0.jpeg

2026-06-24 06:16:15 UTC

Running frontier AI models locally is about to get real.
AMD's Ryzen AI Max+ PRO 495 is landing Q3 2026 with 192GB unified memory per node. Two of these give you 384GB shared CPU/GPU memory — enough to run GLM 5.2 (744B params, 40B active MoE) at 1-2 bit quantization entirely on your desk.
No cloud GPUs. No API credits. No uploading sensitive docs to OpenAI.
The setup:
- 2× PRO 495 nodes (~$9-10K total)
- 10GbE networking between nodes
- llamacpp with RPC for multi-node inference
- 384GB total memory pool
What fits:
- GLM 5.2 at 1-bit (223GB) ✓
- GLM 5.2 at 2-bit (245GB) ✓
- Future 1T+ parameter models with aggressive quant ✓
Why it's future-proof:
Unified memory architecture scales with model size. As open models keep growing, you're not buying new GPUs with fixed VRAM — you're allocating from a massive shared pool. 2 nodes today, add a 3rd later if needed.
The economics:
One-time capex vs $20-50K/year in cloud GPU rental. Pays for itself in 6-12 months for heavy users. And you own the hardware.
Cloud AI isn't going away. But for private agents, sensitive data, and uncensored local inference — this is the inflection point.
#LocalAI #AMD #StrixHalo #GLM5 #OpenSourceAI #AIInfrastructure #DataSovereignty

2026-06-23 20:51:37 UTC

is #codeberg down?

2026-06-23 12:05:52 UTC

man I hate that #codeberg randomly blocks #mullvad nodes.

2026-06-23 10:38:37 UTC

testing #notedeck - was a little bit of work necessary to make it run smooth on kde plasma - ubuntu. didn't pick up the icons etc, claude fixed it

2026-06-23 10:22:00 UTC

@npub1xts…kk5s would be cool if notedeck was available on flatpak 🥺

2026-06-23 10:04:23 UTC

- reply

Yes, I moved to omp.sh which essentially does it via orchestrate

2026-06-22 10:18:15 UTC

- reply

ah nice. will follow and get it once implemented

2026-06-22 10:16:48 UTC

Scored F (15%) for AI agent-readiness on metaend Grade. Scan your site → https://metaend-grade.fly.dev/r?host=noornote.app&grade=F&score=15
@npub175n…g6w0 👋

2026-06-22 10:15:45 UTC

- reply

i need to check out your app. it's paid only atm, isn't it?

2026-06-22 10:13:09 UTC

- reply

moved to codeberg (for private stuff) and gitworkshop for everything else arleady :)

2026-06-22 06:30:39 UTC

DO👏NOT👏KYC👏FOR👏FABLE👏
Either use multi providers like
https://nano-gpt.com/subscription/kUGJhgF9
https://ppq.ai/invite/9bc57632
https://venice.ai/chat?ref=ksiwS7
or get GLM subscription
#nevent1q…r2j9

2026-06-22 05:21:37 UTC

Fuck em not gonna happen
#nevent1q…8ac0

2026-06-21 14:56:34 UTC

- reply

Ahh. Do you know who's the dev here?

2026-06-21 14:38:48 UTC

- reply

But keychat is your app too? Sorry for the confusion

2026-06-21 13:40:01 UTC

- reply

But keychat has its own separate encryption model, doesn't it?

2026-06-21 13:10:38 UTC

- reply

"Post-quantum secure" means the encryption still holds once large quantum computers exist. The catch: a quantum computer running Shor's algorithm breaks the classical key exchange almost every secure messenger relies on, including the X25519 in Signal's X3DH handshake that Keychat is built on.
It matters now, not later, because of "harvest now, decrypt later." An adversary records your encrypted Nostr traffic today and sits on it until quantum computers can crack it. Anything private sent through Keychat this year could be opened years from now. For a messenger whose whole pitch is confidentiality, that is the real exposure.
So the actual question is: does Keychat's Signal handshake use classical X3DH (X25519 only), or the post-quantum hybrid PQXDH (X25519 plus ML-KEM/Kyber) that Signal shipped in 2023? If it is stock Double Ratchet with classical X3DH, then no, not PQ-secure yet, same as most apps. Same question for your MLS group ciphersuite, and note the Nostr secp256k1 identity keys are classical too.
The fix is the hybrid path: keep X25519, add ML-KEM, so you stay secure even if one side breaks. That is what PQXDH and the post-quantum MLS ciphersuites are for.

2026-06-21 11:47:58 UTC

Can you run a #relay in wasm/browser? That would be kinda dope

2026-06-21 10:13:16 UTC

- reply

Sounds pretty cool?

2026-06-21 09:25:42 UTC

- reply

cc @npub175n…g6w0

2026-06-21 07:19:19 UTC

Is #keychat post quantum secure? 🤔

2026-06-20 18:37:21 UTC

Testing #keychat and #noornote on mobile. This is actually working great. Was looking for an app like this!

2026-06-20 13:02:50 UTC

Cutting edge privacy stuff you can use now for cheap:
VPN: nymvpn blockchain based, 1 month free via https://nym.com/pricing?ref=31n6wmEUq62
AI: TEE with local encryption via PPQ https://ppq.ai/invite/9bc57632
Analytics: Cookieless, differentially-private, self-hosted analytics for Astro. https://gitworkshop.dev/[email protected]/relay.ngit.dev/astro-private-count
#privacy #ai #vpn

2026-06-20 12:10:38 UTC

NVIDIA AI Introduce SpatialClaw: A Training-Free Agent That Treats Code as the Action Interface for Spatial Reasoning
https://www.marktechpost.com/2026/06/19/nvidia-ai-introduce-spatialclaw-a-training-free-agent-that-treats-code-as-the-action-interface-for-spatial-reasoning/

2026-06-20 11:09:10 UTC

- reply

Depends on the tools, mine us kinda heavy too 🤔

2026-06-20 10:53:24 UTC

- reply

For omp you need around 47k ctx to be somewhat productive. 30k go for tools and system prompts

2026-06-19 19:02:18 UTC

nostr-web-comment-extension is a reader-side Chrome extension that turns any open webpage into a Nostr discussion thread. When you land on a page and explicitly invoke the extension, it queries Nostr relays for comments anchored to that URL, renders them ranked by your own Web of Trust, and lets you reply or start the thread yourself.
https://gitworkshop.dev/[email protected]/relay.ngit.dev/nostr-web-comment-extension
#ditto #wot #weboftrust

2026-06-19 12:58:31 UTC

"Should we rewrite our CLI tools in Clojure to save on AI token costs?"
I measured it instead of guessing.
We run small command-line tools that our AI agents call constantly. The question was simple: would porting them from Python to babashka (Clojure) make them more economical for an agent to use?
So I ported our flagship tool 1:1, confirmed the output was byte-identical, and then counted tokens and time on both sides.
What I found:
- Tool OUTPUT — the cost that repeats on every single invocation — was identical. 232 vs 230 tokens, byte-for-byte. It is just text; the language cannot change that.
- Source code was about 7% fewer tokens (and 28% fewer lines) in Clojure for the same behavior — but that is a one-time cost you only pay when you edit the tool, not when you run it.
- Cold start: babashka was about 2.5x faster (5ms vs 14ms).
- Real-world run: a wash. The network dominated both by a wide margin.
Verdict: stay on Python.
The lesson generalizes beyond this one choice. When you optimize what an AI agent costs you, optimize the part that REPEATS — the output it ingests on every call — not the language the tool happens to be written in. The implementation language is almost noise. The interface is everything.
Measure before you migrate.
#AI #LLM #AIAgents #Clojure #Python #babashka #DeveloperTools #SoftwareEngineering

2026-06-19 10:01:21 UTC

# nostr-web-comment-extension
**A decentralized, Web-of-Trust-filtered comment layer for any URL, built on Nostr.**
## What it is
`nostr-web-comment-extension` is a reader-side Chrome extension that turns any open
webpage into a Nostr discussion thread. When you land on a page and explicitly invoke
the extension, it queries Nostr relays for comments anchored to that URL, renders them
ranked by your own Web of Trust, and lets you reply or start the thread yourself.
It is built entirely on existing, standardized primitives — **NIP-73** assigns every
URL a canonical identifier, and **NIP-22** (kind `1111`) threads comments onto that
anchor. This is the same mechanism [Ditto](https://soapbox.pub/ditto/) already uses,
so a corpus of URL-anchored discussion exists today and this extension is interoperable
with it on day one. There is **no new protocol to invent, no central server to operate,
and no cooperation required from the site owner**.
https://gitworkshop.dev/[email protected]/relay.ngit.dev/nostr-web-comment-extension
#nostr #ditto #dev #opensource
https://blossom.primal.net/4d665e860d9cd2b49db2e8c02cd225d1e6ef53ed4433c6cefe198b5c31d903c3.jpg
https://blossom-01.uid.ovh/09831e42acaf81b26b72232a9b5d56d233dd3d3b84842e3ea5f2005c1b9364ff.jpeg

2026-06-19 07:19:24 UTC

building Reader-side #Nostr (NIP-73/NIP-22) comment layer for any URL, Web-of-Trust filtered.

2026-06-19 06:28:24 UTC

Fable 5 is back?

2026-06-19 05:29:49 UTC

- reply

Cool thanks! Might vibe my own tho

2026-06-18 19:10:39 UTC

Introducing eve, an open-source agent framework
https://vercel.com/changelog/introducing-eve-an-open-source-agent-framework

2026-06-18 13:35:12 UTC

Is there a #ditto extension that lets me directly comment on an open URL AND shows comments? Would be a cool WoT

2026-06-18 13:27:27 UTC

Kinda cool 🤔
Solar panels are creating a strange effect by forming rainfall clouds and thriving oases in the middle of the desert
https://www.ecoportal.net/en/solar-panels-strange-effect-are-creating-rain-clouds/19854/
#solar

2026-06-18 11:14:40 UTC

- reply

what the fuck?

2026-06-18 09:23:53 UTC

is such a cool publishing platform for atproto, looks very complete and useful. Is there something remotely similar on nostr?
https://offprint.app/

2026-06-18 05:25:42 UTC

- reply

Ethereum Name Service 👋