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 17:29:18 UTC

Built a fully local ai system with LFM2.5 as harness and gemma4 E2B as agent. I want to do some research on how to make it work for banking/crypto

2026-07-10 16:07:12 UTC

Sigh another polymarket round for the insiders
#nevent1q…vd4g

2026-07-07 15:13:48 UTC

- reply

Just use /advisor 👋

2026-07-02 15:16:07 UTC

- reply

Yes, I built this https://gitworkshop.dev/ngmi@zaps.lol/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

2026-06-29 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

2026-06-28 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

2026-06-27 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 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

2026-06-26 12:22:13 UTC

- reply

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

2026-06-26 05:25:52 UTC

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

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

2026-06-24 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/ngmi@zaps.lol/relay.ngit.dev/nostr-web-comment-extension

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 05:21:37 UTC

Fuck em not gonna happen
#nevent1q…8ac0

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 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 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 07:19:24 UTC

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

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 05:25:42 UTC

- reply

Ethereum Name Service 👋

2026-06-17 19:22:08 UTC

Cool? Is this the ens for monero?
https://xns.rocks/
#monero #xns

2026-06-17 07:35:41 UTC

- reply

But is this post quantum resistant?

2026-06-16 18:40:38 UTC

- reply

Wen playstore?

2026-06-16 18:30:15 UTC

- reply

Yes

2026-06-16 18:27:59 UTC

Why does #ditto not rich format markdown? Weird.

2026-06-16 18:10:40 UTC

- reply

Mobile browser apps to use are kinda whack

2026-06-16 13:26:36 UTC

Opus and me are helping to improve #omp for local models
https://github.com/can1357/oh-my-pi/issues/2774#issuecomment-4718424478
issue closed and fixed already!

2026-06-16 07:06:40 UTC

- reply

Found a dead link https://satsigner.com/ doesn't load here

2026-06-15 09:07:48 UTC

- reply

the qwen-0.6b intruct models run literally anywhere, which is cool too

2026-06-15 05:22:36 UTC

- reply

Yep

2026-06-14 19:18:41 UTC

- reply

Yeah mullvad requires systemd for some reason

2026-06-14 13:41:13 UTC

Scored F (25%) for AI agent-readiness on metaend Grade. Scan your site → https://metaend-grade.fly.dev/r?host=switchboardgo.com&grade=F&score=25
Might want to make this more accessible for agents. 👋
#nevent1q…f9ne

2026-06-14 08:43:30 UTC

@nprofile…a7q0 is this true? for Mullvad (its .deb postinst calls systemctl and fails to install on Devuan

2026-06-14 08:39:38 UTC

Moving to systemd free #Devuan soon. Initially I had #artix in mind, but the ecosystem is such a cluster fuck... https://www.phoronix.com/news/Arch-Linux-AUR-More-Than-1500

2026-06-14 07:40:24 UTC

BREAKING: NEW CLAUDE PLAN
/s
#Claude https://blossom.primal.net/42742c0d93a2a6db8bf753c11f74c0fc9b7bc60908e0a98f6bcc80c5837f060f.jpg

2026-06-14 06:20:51 UTC

- reply

Not sure what you mean?

2026-06-13 15:58:31 UTC

- reply

Better to be prepared 👏

2026-06-13 13:23:09 UTC

First time I had to use #bitchat. In a crowded big mall with dead Internet, to stay in contact with my partner after split. I wish other messengers like #whitenoise abd @nprofile…vl4s would invest energy on adding BLE and meshtastic compatibility

2026-06-13 05:44:01 UTC

Ditching proprietary lock-in. Moving the workflow from Fable to open source weights using oh-my-pi (omp).
The setup strategy:
- Plan: Kimi K2.6 for orchestration and architecture
- Slow: Kimi K2.7 Code for deep implementation
- Smol: Gemma 4 for fast iterative edits
You can configure these roles directly in the TUI or override at launch with flags like --plan or --slow. For a subscription experience without the API credit headache, Venice and Moonshot are the way to go.
APIs:
Sub/Pay with crypto:
🪡 https://nano-gpt.com/subscription/kUGJhgF9
🪡 https://ppq.ai/invite/9bc57632
Stake $VVV for permanent daily tokens:
🪡 https://venice.ai/chat?ref=ksiwS7
#AI #OpenSource #CodingAgent #ohmypi #Kimi #VeniceAI #DeveloperTools #venice

2026-06-12 11:07:59 UTC

- reply

link?

2026-06-11 14:37:11 UTC

- reply

Lmao same for the tokens. How do you like omp? I recently switched, and I love it

2026-06-10 07:14:56 UTC

- reply

But how is ut compared to opus xhigh? I haven't had any problems with opus at all so far. So I wonder if it's even worth it to get used to fable just to get rugged in 2/weeks

2026-06-09 12:33:03 UTC

- reply

do they take monero as investment?

2026-06-09 06:53:40 UTC

But how...? Isn't WA supposedly E2EE?
#nevent1q…8n8y

2026-06-09 06:50:43 UTC

- reply

Found a pinned comment in the proton sub (still doesn't look good imo) https://www.reddit.com/r/ProtonMail/s/4cexnrP40P

2026-06-09 06:00:31 UTC

- reply

I've got a fracture in a finger rn, was never black. Could be both under the bruise tho

2026-06-09 05:00:34 UTC

- reply

Difficult to scan a qr with the same device you wa t to login with😀

2026-06-08 18:40:15 UTC

What the fuck
https://www.reddit.com/r/BuyFromEU/s/vfcKpuhbJp
https://blossom.primal.net/520e449cd76d225581fefccb17e1fa1b664325d47c5d5143ac560a2f4c271923.jpg
#protonmail

2026-06-08 10:41:48 UTC

I Built Analytics That Cannot See You
https://paragraph.com/@metaend/analytics-that-cannot-see-you
#DifferentialPrivacy #PrivacyByDesign #OpenSource #Astro #PrivacyEngineering #WebDevelopment

2026-06-08 07:49:34 UTC

Does damus offer bunkr remote login functionality? Can I use it to login to webapps?
@nprofile…vgs2

2026-06-08 07:03:01 UTC

The biggest threat to censorship resistance are humans being assholes

2026-06-08 05:37:27 UTC

Is there a playstore app that supports bunkr and whatever else is needed to use nostr in android? I wish all mobile clients would just support it

2026-06-07 17:55:23 UTC

Is there a EUR sats backed version of cashu? If yes, is there also interest? Or too primitive

2026-06-07 16:33:07 UTC

- reply

What's the app?

2026-06-07 08:02:05 UTC

astro-private-count - cookieless, differentially private. live on https://metaend-grade.fly.dev/
https://metaend-grade.fly.dev/stats
git: https://gitworkshop.dev/ngmi@zaps.lol/astro-private-count
"Counts are differentially-private estimates, not exact tallies. Each visitor's yes/no bit is randomized in their browser (epsilon = 1.099, so a report is truthful 75.0% of the time) and totals are de-biased here. Rows with under 100 reports are noisy and shown with a ~. This is local differential privacy — the server learns totals and nothing about any individual. It is not a zero-knowledge proof and does not anonymize the network connection."
#astro #privacy

2026-06-07 06:54:41 UTC

PROMPT :
Draw me as if an obsessed fan artist filled an entire sketchbook page - messy, overlapping, full-body poses, tiny chibi doodles, exaggerated expressions, and random close-ups of their hands or eyes. White background. No grid, no order. Pure chaos energy
Via nano banana 2
https://nano-gpt.com/subscription/kUGJhgF9
https://ppq.ai/invite/9bc57632
https://blossom.primal.net/d847682ffaee886791edd5babe8c35854734a033342ecdf1b7746d5a29cbb221.jpg

2026-06-07 06:18:22 UTC

- reply

Following you now 👋

2026-06-07 05:39:49 UTC

Solana deposits now also support USDT. You can choose USDT on Solana from the deposit flow, get a Solana payment link for the selected asset, and fund your NanoGPT balance through the same Solana deposit option.
https://nano-gpt.com/subscription/kUGJhgF9
#ai #solana

2026-06-06 05:55:54 UTC

Running gemma 4 via unsloth locally literally replaces my usage of smaller 3rd party TEE models. Need to find an uncensored version tho
#gemma #unsloth

2026-06-05 13:37:01 UTC

Isle Goblin Gameplay Trailer
https://www.youtube.com/watch?v=6lAYtIFfsGA

2026-06-05 13:21:13 UTC

this mashup tho
https://soundcloud.com/user505229234/linkin-park-slipknot-eminem-till-the-end-mashup

2026-06-05 12:19:49 UTC

Oof
#nevent1q…zz04

2026-06-05 10:48:57 UTC

wait wat
oh my pi CLI using 38% of my CPU 😂
https://blossom.primal.net/5b0629659a4c38a3de9b3e9caa2b18ef1be01d4e13feb731c406bc2d35a2808c.png

2026-06-05 07:42:40 UTC

Heh
#nevent1q…2xlw

2026-06-04 10:39:30 UTC

- reply

The recovery gap is real. The relay-trust gap is a category error.
A relay is a public transport layer, not your database. Storing private data on a random relay and complaining it is public is running prod on a pastebin. Nostr does open identity and broadcast; private data, backups, and recovery belong on an encrypted layer underneath, which exists.
And a password reset link is just a custodian you forgot you trusted. The gap is UX maturity, not structural.

2026-06-04 10:10:50 UTC

The Meta AI Hack Is a DISASTER
https://www.youtube.com/watch?v=wlMgNtBipe4

2026-06-03 18:42:37 UTC

This fish species survived 100,000 years without males. Scientists thought it should be long dead – but it's thriving
https://www.bbc.com/future/article/20260601-amazon-molly-the-fish-thats-lived-for-100000-years-without-males

2026-06-03 17:59:24 UTC

Scored F (25%) for AI agent-readiness on metaend Grade. Scan your site → https://metaend-grade.fly.dev/r?host=hodlhodl.com&grade=F&score=25
@nprofile…05yu

2026-06-03 14:36:10 UTC

Scored F (30%) for AI agent-readiness on metaend Grade. Scan your site → https://metaend-grade.fly.dev/r?host=zeusln.com&grade=F&score=30
@nprofile…2xhz

2026-06-02 17:49:22 UTC

- reply

Finally someone treating the Nostr key as the identity instead of bolting a DID registry back on top. "No trusted setup, runs on an ESP32" is the whole pitch in one line. The thread I keep pulling: agents already carry Npubs, so the same credential plus revocation plus ZK layer becomes delegated-authority proof for autonomous agents, not just humans at a turnstile. Prove an agent is authorized to act without doxxing the principal behind it. Is agent authorization in scope for you, or are you keeping it human-first for now? Reading the protocol this week either way.

2026-06-02 02:41:24 UTC

Crazy, notes my agent writes are not visible in amethyst but on ditto. Is this a relay thing?

2026-06-01 08:26:56 UTC

really?
The public #bounty tag is a cesspool. Top results were CSAM/illegal-content spam ("Cp jk", "Teens", "Rape"... Telegram links).
The naive tag-fetch ingests this. Two real issues:
1. We'd be fetching and persisting that text onto your disk (pipeline state, git-crypted repo, audit log). Legal/hygiene hazard.
2. The only filter today is the feasibility LLM downstream — that content should never enter the pipeline at all.

2026-06-01 07:43:27 UTC

- reply

i vibed this https://metadm.shakespeare.wtf and it works very good with amethyst, that's why i am confused. dm on nostr is really messy

2026-05-31 19:42:02 UTC

Oh my pi looks sick! I have to test it.
https://omp.sh/

2026-05-31 17:31:37 UTC

The Agent Credential Problem Just Got a Real Answer
https://paragraph.com/@metaend/agent-credential-problem-just-got-a-real-answer
#agents

2026-05-31 17:00:57 UTC

@npub1pls…kx8j
Scored F (25%) for AI agent-readiness on metaend Grade. Scan your site → https://metaend-grade.fly.dev/r?host=openvibe.social&grade=F&score=25

2026-05-31 16:56:18 UTC

Scored F (5%) for AI agent-readiness on metaend Grade. Scan your site → https://metaend-grade.fly.dev/r?host=bitcoinist.app&grade=F&score=5
@npub174w…vajv

2026-05-31 16:50:21 UTC

Why am I not seeing the DMs on @npub17n4…fyh9 but only on @npub142g…xrj0 ?

2026-05-31 08:59:10 UTC

Venice has a large amount of TEE models included in their subscription, total worth it if you use general AI (no api/coding) IMO
https://venice.ai/chat?ref=ksiwS7
https://blossom.primal.net/1d5458747a16c97d26a9c7595c98ff1f2aa40f6258fa864865f757a1b94068ff.jpg

2026-05-31 08:43:04 UTC

- reply

Sounds super fucking cool. Woudl love to read a in-depth writeup on it 🙏

2026-05-31 08:06:51 UTC

Sign up at uomi and you get $3.50 worth of computation for free to use qwen, DeepSeek, minimax chat&api. It's a great offer, and I use it with my hermes as secondary
https://uomirouter.uomi.ai/auth/signup?ref=XZG8J3E8

2026-05-30 15:17:23 UTC

Microsoft just shipped the Agent Governance Toolkit. My read: validation, not competition.
AGT wraps policy enforcement around agents and evaluates every action before it runs, checking against your policies, allowing or denying, and logging the result. It covers the OWASP Agentic Top 10, supports intent-based authorization, token budgets, kill switches, and a tamper-evident decision record. Sub-millisecond per operation. Solid work.
If that thesis sounds familiar, it should. I have been building WasmBox toward the same core idea for months: agents are a new class of user, and you do not control them with rules in a prompt. You control them at the boundary, before the call executes. Seeing the largest software company on earth converge on the same architecture is a good signal.
Where the two diverge is the isolation boundary.
AGT is candid about this in its own README: it enforces governance at the application middleware layer, in the same process as the agent, and recommends running each agent in a separate container for OS-level isolation. Reasonable choice. It also leaves the strongest control off the critical path.
WasmBox draws the boundary at the runtime. Two layers:
Sandboxed tools run as isolated wasm32-wasip2 components with zero capabilities by default. No filesystem, no network, no env unless explicitly granted, enforced by Wasmtime, not by policy living next to the code it governs. Network access is host-filtered: a tool can only reach the hosts declared in its manifest, and every other request is rejected before it leaves the sandbox.
Everything else the agent runs gets recorded. The 0.4 session recorder proxies every native command through a thin wrapper (wasmbox via), generates PATH shims, and launches the agent in a recorded environment (wasmbox claude, or wasmbox wrap for any agent). Every command is logged with binary path, SHA-256 hash, arguments, exit code, and a session ID. You can export a signed, Ed25519-stamped audit report for a single agent session.
So the split is clean: sandboxed WASI tools give you isolation, the proxy gives you visibility, and the signed log gives you provable evidence. Built for EU AI Act Article 14 from the start, not retrofitted.
Both projects share the same philosophy: treat the agent as untrusted code, isolate it, govern it at the edge. The difference is where you draw the trust boundary. I think the runtime is the right place to draw it, and I am glad to see the industry moving in this direction.
WasmBox is at v0.4.1. Ten tools live in the demo registry, single binary, no accounts, no telemetry. Install is one curl line.
https://tangled.org/metaend.eth.xyz/wasmbox-cli
Curious how others are thinking about this: process-level governance plus containers, or capability isolation at the runtime? Where are you drawing the boundary?
#AIAgents #AgenticAI #WebAssembly #WASM #AISecurity #AIGovernance #EUAIAct #Rust #AgentOps #OWASP #RuntimeSecurity #Sandboxing #AICompliance #OpenSource

2026-05-30 10:39:42 UTC

- reply

You need relays for it to work. So there is already a dm relay list inside

2026-05-30 07:57:34 UTC

- reply

Just built a decentralized customer chat for my projects, fully encrypted via nostr. This is the way 👋https://metadm.shakespeare.wtf/

2026-05-29 20:03:54 UTC

Europe passed the GDPR and now wants to pass the AGI Act. The original sin is the same: regulate what you don't understand until nobody can build it. Startups don't even need to go to Silicon Valley anymore. Dubai, Singapore, wherever there's a power outlet and no ministry of compliance waiting with a fine.
#europe #startups #regulation

2026-05-29 17:58:39 UTC

- reply

Cool!

2026-05-29 09:41:44 UTC

Scored D (45%) for AI agent-readiness on metaend Grade. Scan your site → https://metaend-grade.fly.dev/r?host=ppq.ai&grade=D&score=45
@npub16g4…kv4h :aim-cool: