ngmi

https://metaend.qstorage.quilibrium.com

hi~

Public Key

npub14p7fhlz9wrlptt7a5mqqjkk79460k98r999l7dafv4fznqaa565qhjdzry

NIP-05 Address

Profile Code

nprofile1qqs2slyml3zhpls44lw6dsqftt0z6a8mzn3jjjllx75k253fsw76d2qpz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0ds2m208c

Publishing to

relay.damus.io

nos.lol

Author Public Key

npub14p7fhlz9wrlptt7a5mqqjkk79460k98r999l7dafv4fznqaa565qhjdzry

Show more details

Last Notes

2026-11-11 14:55:42 UTC

this shit is the norm with @npub168e…9uaw nowadays. had to write a omarchy widget to have a chance to at least find a somewhat working exit node ffs
https://metaend.blossom.band/f344762ff0265964fb45c18076c02d83ff4857bff5b590187d3943c2a9a5c36a.png

2026-11-11 13:10:42 UTC

Creating a little idle rpg game in #herdr with my agents
https://metaend.blossom.band/fab73a4f474e22fbd551544673541248cb28069e5d00d3495c65bb8174e24c5e.png

2026-11-11 12:02:24 UTC

When PII redaction is on for a zero data retention request, NanoGPT now always redacts it with Tinfoil, even if Grepture was chosen in the request, the API key or your settings, so the request never passes through Grepture. Detection runs in Tinfoil's attested, encrypted enclave. A request requires zero data retention when its API key is set to it (or the key has no setting and your account is), when it is billed to a team set to zero data retention, or when it sends provider.zdr: true. Zero data retention doesn't turn redaction on by itself. Tinfoil's redaction pricing applies.
https://nano-gpt.com/r/qzFYzwRg

2026-11-10 17:47:51 UTC

Eight AI decision models traded one day of Bitcoin. Doing nothing won.
https://paragraph.com/@metaend/eight-ai-decision-models-traded-one-day-of-bitcoin-doing
#bitcoin #ai

2026-11-10 16:48:10 UTC

I asked 8 decision models to trade BTC 1h candles, and 7 of them were total losers.
https://metaend.blossom.band/0d2bf25fb3005864ec8afef2564481f493257dd54f9ee5fcf279a1d8c9919fd3.png
Jeff is available on https://nano-gpt.com/r/HCdbQetV
#bitcoin

2026-11-10 13:35:12 UTC

Setting up my #herdr
https://metaend.blossom.band/c4f82d1abeaa412203b3e9f5f8d74c8e6e6efd2269ef962ecdaae2640e3e2a3c.png

2026-11-10 10:20:40 UTC

Read ~/.config/herdr-app/AGENTS.md, then add this mod: https://github.com/mathieucroset/herdr-pokemon-agents/tree/main

2026-11-10 08:51:27 UTC

Our company harness Cowork:Alice RED when it checks a repo and finds secrets in the commits
https://gifverse.net/media/8ATc1QQI/original.gif

2026-11-09 16:21:05 UTC

Tested a 30mb voice model, it's very good! unfortunately non-commercial licence for some reason
https://metaend.blossom.band/2b50deda2af413d19faeebed142508959b8487f7625d21f5d1e2b3c15f677536.mp4
'Voice by Underdog Woof.'

2026-11-09 12:58:04 UTC

is there something like aave but for btc?

2026-11-09 07:38:53 UTC

Jev 1.14 now accepts up to 1,000 inline PNG or JPEG images per Decisions API request, up from four, with up to 3.75 MiB of image data in a request of up to 4 MiB. Each image can be 280 to 1568 pixels per side. A 280-pixel image counts as about 240 input tokens and a 1568-pixel image as about 6,300, and an image request can use up to 262,144 tokens. Images are billed as input tokens at the usual Jev 1.14 rate, and zero data retention works the same way. Requests with moderation remain limited to four images.
https://nano-gpt.com/subscription/kUGJhgF9

2026-11-08 16:38:45 UTC

what's so good about @nprofile…twc2 that makes you pay for it?
Also hello world from #yakihonne

2026-11-08 15:43:55 UTC

Now imagine using heretic version of DeepSeek, should be closer to 90% then
#nevent1q…d5fk

2026-11-08 13:47:12 UTC

Leave those countries if you own crypto and your personal info was leaked. No need to stay where the government fails all the time from leaking to protection
#nevent1q…r86z

2026-11-08 13:24:50 UTC

Giving my #muse access to #nanogpt so it can use jev is actually a massive gigachad move.

2026-11-08 05:50:01 UTC

Armada ultra account owner here, btw

2026-11-08 03:48:23 UTC

LFM3 1.7B is now available. Liquid AI's compact 1.7B-parameter text model reasons before answering and supports tool calling, structured output, and data extraction. It has a 262,144-token context window and up to 8,192 output tokens. Pair it with retrieval for knowledge-heavy tasks; it is not recommended for complex multi-step reasoning or agentic coding.
https://nano-gpt.com/subscription/kUGJhgF9
#lfm #ai

2026-11-07 09:23:15 UTC

https://gifverse.net/media/EQP2Qtr6/original.gif
#nevent1q…j728

2026-11-06 12:12:41 UTC

How the Internet Uncovered a Massive Pokémon Fraud
https://www.youtube.com/watch?v=oc8G53Tbsvs

2026-11-06 08:04:29 UTC

- reply

mullvad has entry and exist nodes you can choose from

2026-11-06 07:47:33 UTC

- reply

no not over tor even, just mullvad - need to avoid to geoblocking. it's ass

2026-11-06 07:23:12 UTC

- reply

yeah i need to use us exit nodes and they are extremely volatile. cant download or upload shit there, switching a lot

2026-11-06 06:20:10 UTC

most of @npub168e…9uaw exist servers are total shit nowadays. so sad to see a once great service go to the garbage

2026-11-06 06:19:05 UTC

most of @npub168e…9uaw exist servers are total shit nowadays. so sad to see a once great service go to the garbage

2026-11-05 08:28:55 UTC

https://gifverse.net/media/Eykb-emH/original.gif
#nevent1q…exhy

2026-11-05 04:53:42 UTC

- reply

yes, love it!

2026-11-05 01:49:31 UTC

🤣🤣
#nevent1q…9mde

2026-11-03 09:40:43 UTC

We're doing free and open security research with our harness
If you want to support, get a cheap $12 subscription at nanogpt via our referral (discounts apply if you pay with BTC or Monero!)
https://nano-gpt.com/subscription/HCdbQetV
https://metaend.blossom.band/305b6fae9d2cf1782cc3241103f9d18913096961ad7b89993e09565f3b2b07d8.png
#bitcoin #monero

2026-11-03 08:31:39 UTC

- reply

pointing abliteration flash and deepseek on the repo now. its still at the 21 version, but i guess thats okay

2026-11-03 01:43:47 UTC

- reply

Not surprising. Google is just bad

2026-11-03 01:42:08 UTC

- reply

Trust me bro benchmark?

2026-11-02 12:30:11 UTC

what kind of xmr exchange is wagyu? just heard of it from zach

2026-11-02 12:04:18 UTC

- reply

yep

2026-11-02 11:55:18 UTC

omarchy-update used to fail a whole transaction on one transient download stall (pacman aborts at 1 byte/10s). Filed it in September, got the caching theory wrong, proved that with ranged GETs, and the thread collected two independent reproductions and a 20 MiB boundary clue. Maintainer traced it to an HTTP/2 flow-control bug in Cloudflare. Closed today. Good thread.
https://github.com/omacom/omarchy/issues/11793
#Omarchy #ArchLinux #OpenSource #Cloudflare

2026-11-02 08:46:30 UTC

- reply

btw how do i collapse large posts again in the web view? taking a lot of space in the tl

2026-11-02 06:44:18 UTC

Happy to see #bitchat getting some traction

2026-11-02 02:40:02 UTC

DeepSeek V4.1 Flash Heretic and its Thinking variant are now included in the NanoGPT subscription at 1x included-input usage. Both are served by Sable and support text and image input, tools, structured output, and optional reasoning across a 512K context window.
https://nano-gpt.com/subscription/kUGJhgF9

2026-11-01 10:35:47 UTC

Checking parser markdown against the PDF, without an LLM
https://paragraph.com/@metaend/checking-parser-markdown-against-the-pdf

2026-11-01 10:03:22 UTC

- reply

freaks come out at night

2026-11-01 09:28:05 UTC

Release: dpp-check
A conformance checker for the Document Parsing Protocol (draft)
https://gitworkshop.dev/c@mailstr.app/relay.ngit.dev/dpp-check
#parsingprotocol #airesearch #markdown

2026-10-31 17:31:31 UTC

- reply

ahhh thanks got it

2026-10-31 17:27:13 UTC

@npub1fa0…0z97 none of these collapse in desktop browser for me
https://metaend.blossom.band/42e1d1ca9a755618bcb95c44e255d6edf227861218b74e8af48d5e1776d0294c.png

2026-10-31 17:23:24 UTC

Just released: Document Parsing Protocol conformance checker: deterministic checks of parser output against the draft protocol (parseprotocol.com)
https://gitworkshop.dev/c@mailstr.app/relay.ngit.dev/dpp-check
#parseprotocol #dpp #research

2026-10-31 06:23:04 UTC

- reply

whats this magic?

2026-10-31 03:40:27 UTC

OrcaCyber Zero 1.5 is OrcaRouter’s cybersecurity model, post-trained for authorized security research: vulnerability research and reproduction, exploit development, penetration testing and security auditing. It accepts text only and supports tool calling, structured output and optional reasoning, with a context window of about 1M tokens and up to 128,000 output tokens. Offensive work such as exploit or payload development is permitted only within a security engagement you are authorized to perform.
It costs $3 per million input tokens ($0.75 cached) and $7.50 per million output tokens, and is pay-as-you-go only, not included in the subscription.
https://nano-gpt.com/subscription/kUGJhgF9

2026-10-30 10:10:16 UTC

- reply

hiii

2026-10-30 10:07:03 UTC

genius, but now aave is royally fucked
#nevent1q…vhth

2026-10-30 08:36:49 UTC

are they completely out of their fucking mind. what are these processes to allow one laptop to control the whole fucking system
#nevent1q…mywj

2026-10-30 03:08:56 UTC

Bye-nance

2026-10-29 10:57:28 UTC

You can now limit where your API requests run. Turn on EU only under EU inference for API keys to apply it to all your API keys, or set Inference region in a single key's settings. Keys follow the account setting unless they set their own region.
With EU inference on, the first route and every fallback must be a reviewed EU route for the model you request. If no EU route can serve a request, it fails with eu_inference_route_unavailable instead of running elsewhere. Most routes run in EU member states; some models run on European endpoints that can also be in the EEA, the UK or Switzerland. With Zero Data Retention also on, a route must meet both.
More than 100 models have an EU route, including Claude, Gemini Flash Lite, Mistral, DeepSeek V4.1 Flash, GLM 5.3, Kimi and MiniMax M3. The setting lists every model and where its routes run. EU inference works with the Chat Completions, Messages and Responses APIs. GPT models, Auto Model, Decisions models, batches, web search, URL scraping, memory, PII redaction, your own provider keys, legacy /v1/completions, and image, video, audio and embedding endpoints are not available with it.
EU requests count as choosing a provider: they are pay-as-you-go, not included in a subscription, and billed at the EU route's listed price plus the 5% provider-selection fee.
The setting covers where models run. Requests still pass through NanoGPT's servers, and your account, billing, support and stored data are not moved to the EU. Web chat is not covered.
https://nano-gpt.com/subscription/kUGJhgF9

2026-10-28 12:39:11 UTC

@npub1fa0…0z97 i see notifcations in web, awesome! only thing is the badge doesn't go away after i checked all :)

2026-10-28 09:18:50 UTC

Rune 26B-A4B is now available for classification, scoring and yes/no judgments. Ask up to 256 typed questions in one request and receive calibrated probabilities. Rune accepts text and JSON, supports up to 208 choice options and 2–10 score levels, and generates no conversational text or reasoning. Use /v1/decisions or /v1/systemone. Input costs $0.042 per million tokens, cached input $0.008 per million tokens, and output is free.
https://nano-gpt.com/subscription/kUGJhgF9

2026-10-28 07:36:38 UTC

but these announcements are not new, most of them are old
#nevent1q…w8xc

2026-10-28 05:47:22 UTC

- reply

just hope he won't become a gemini clanker

2026-10-28 05:41:54 UTC

Every platform now ships an agent that works for you 24/7. Meta has Muse, and the rest all have theirs.
Buyers use them. They ask one question instead of opening ten tabs.
Those agents answer from your website. The website you built for humans. Pages that contradict each other, prices that disagree, hours that don't match the footer.
An assistant doesn't ask you to clarify. When it gets two answers it picks one. When it can't describe you clearly, it describes a competitor it can.
SEO got you found by people. GEO is getting you answered correctly by assistants.
I run a free AI Readiness Score: 40 buyer questions across ChatGPT, Gemini, Claude and Perplexity. One page: where you show up, where competitors do, what to fix first.
Reply with your site and I'll run yours.

2026-10-27 09:34:30 UTC

- reply

nostrvault is neat! only thing I'm missing (?) is notification badge unseen. unless I miss this setting somehow

2026-10-27 06:58:22 UTC

at the #DubaiAIFestival and I'm quite certain that 95% of the companies here won't exist in a year max two. so many wrappers and very little creativity

2026-10-26 18:12:52 UTC

that's pretty cool!
#nevent1q…6827

2026-10-26 17:12:21 UTC

how's the new cyberpunk?

2026-10-26 13:49:08 UTC

Your AI coding agent obeys whatever AGENTS.md sits in the folder you open. A web page can drop a file into your Downloads folder silently, no click, no prompt.
Forever Security caught exactly that: a planted AGENTS.md in Downloads told the agent to install a malware-linked npm package. Codex ran it in over 90% of tests. A Claude Code variant that redirected API traffic to steal the session token worked about half the time.
The fix is boring and local: never start an agent in ~/Downloads, read instruction files you did not write, and check your agent config for servers you did not add.
Trust in the file is the feature. That is also the hole.
https://thehackernews.com/2026/10/malicious-file-in-downloads-folder.html
#ai #aisec

2026-10-25 15:39:27 UTC

My muse is dressed now for Halloween 🎃
https://blossom.ditto.pub/f0f53fed827e80c85f43705c2801616f4f57a34cb82f6e15f765f722aa273c42.png

2026-10-25 13:17:03 UTC

Gm
https://blossom.ditto.pub/ddd868371fd5724f67fff2fcc7ef8a42257a969867a1c876160714d88bf9f171.png

2026-10-25 12:09:35 UTC

- reply

opus doesn't even do simple security audits for me without showing me the finger :(

2026-10-25 12:07:05 UTC

- reply

fully agree, and you can audit for a couple of cents or dollars depending on the size. I use rn abliteration-ai/abliterated-model-large-flash which is deepseek latest abliterated, so it won't tell me to fuck off because SEHCUHRETEH like claude for example

2026-10-25 03:57:10 UTC

getting a lot of haven sync issues in nostrvault tho. app looks cool, but seems to have issues

2026-10-23 17:23:27 UTC

Abliterated Large Flash is now available! An uncensored reasoning model based on DeepSeek V4.1-Flash-1017, it accepts text and images and returns text. It supports tool calling, JSON mode, JSON Schema output, and automatic prompt caching across a shared 1,048,576-token context, with up to 262,144 output tokens. Reasoning defaults to high, with off, low, high, and max modes.
https://nano-gpt.com/subscription/HCdbQetV
#deepseek #secops #ai #red

2026-10-23 03:28:57 UTC

Synthetic launched a $9 go plan with 1 billion DeepSeek tokens per month. That's huge!
Also infinite mode for always on agents that's slower, but doesn't count against token usage!
https://synthetic.new/?referral=ndEXRFUQ
#deepseek

2026-10-22 04:35:20 UTC

Gm 🤣 https://metaend.blossom.band/63eb71e3aed89ed139bf17f085623985ed8849b32cda91578f902ab156e09b9d.mp4

2026-10-21 15:19:15 UTC

Certifi deez nuts

2026-10-21 09:30:04 UTC

Hackers Use Autonomous AI Agents to Breach Company Websites, Target Over 1,200 Sites
https://thehackernews.com/2026/10/hackers-use-autonomous-ai-agents-to.html?m=1

2026-10-21 09:17:53 UTC

- reply

a human intelligence?

2026-10-21 08:17:56 UTC

Who would've thought REA would become Satoshi's bane
#nevent1q…kkrx

2026-10-21 06:37:08 UTC

Testing orca, it's a bit clunky (not a fan of ui managers) but right now really helpful since I have a lot of projects with agents running.

2026-10-19 18:09:11 UTC

Most decision models are now included in your subscription: Jev 1.13, Jev 1.14, Jev Latest, Decider 0.8B FP8, Decider 2B FP8, Decider 4B NVFP4, Liquid D1, PPLX Decider v1.1 27B, Kev 4B TEE, SubQ Snap Preview, eu/jev Jeff 1.0.0, and eu/jev Jeff Latest. Their input tokens count toward your included input tokens at the regular 1x multiplier through POST /api/v1/decisions, POST /api/v1/systemone, and the text-only questions format in Chat Completions, Messages, and Responses. API keys set to subscription-only billing can use them too.
Clef, Clef Flash, Clef Omni, and GPT-6 Luna Decisions remain pay-as-you-go, and PII redaction is still charged separately.
https://nano-gpt.com/subscription/kUGJhgF9

2026-10-19 07:30:50 UTC

#nevent1q…0caz
https://static.klipy.com/ii/d7aec6f6f171607374b2065c836f92f4/39/8d/82x0RGji.gif

2026-10-18 08:14:16 UTC

Perplexity Open-Sources SPACE, a Firecracker MicroVM Sandbox for AI Agents
https://www.marktechpost.com/2026/10/16/perplexity-open-sources-space-firecracker-microvm-sandbox-ai-agents/?amp=1

2026-10-17 13:13:05 UTC

https://gifverse.net/media/nF7D56cP/original.gif
#nevent1q…jsrs

2026-10-17 10:07:50 UTC

What the fuck lmao. Farcaster has become so cringe
https://blossom.ditto.pub/38e8141e1852e9b788826ab8dd85bfe532950c19978cd50e2233d9c1a302ebcc.jpg
#farcaster

2026-10-17 09:53:12 UTC

This is kinda crazy. I gave muse an api key for wan, now it generates videos based on knowledge for me. All in mobile browser and "free".
1 billion tokens for free (use US in vpn): Check out Muse, your personal AI agent. Redeem my code in Settings within 48 hours of joining and we'll both get 1 billion Muse tokens.
Code: 8WQ4S7
https://muse.ai/join

2026-10-17 05:05:02 UTC

eu/jev Jeff is now available for classification, routing, scoring, and yes/no decisions. Send text or JSON state with typed Choice, Score, or Noul questions and receive structured answers with probabilities. Inference runs in Germany, with a 32,768-token context and up to eight questions per request.
https://nano-gpt.com/subscription/kUGJhgF9

2026-10-16 07:53:43 UTC

My work "Shitposting on nostr" scored 79/100 on AI exposure (Very high) with the free Tideproof scan. How exposed is yours? https://tideproof.org/score/very-high/?utm_source=nostr&utm_medium=social&utm_campaign=scan_share
fuck 🫡

2026-10-15 17:05:06 UTC

- reply

https://static.klipy.com/ii/4493325008d34b7bf8cd6813cd5c1619/50/53/yB3pEk31iXQYm.gif

2026-10-15 06:13:12 UTC

Jev 1.14 is now available for routing, classification, scoring, and typed yes/no judgments, with improved multilingual performance. It returns typed answers with probabilities instead of generated text. Use typesafe/jev-1.14 through the Decisions API, TypeSafe SDK compatibility, or a questions response format through Chat Completions, Messages, or Responses. This API-only release supports text and JSON input within the default context limits.
https://nano-gpt.com/subscription/kUGJhgF9

2026-10-15 05:00:24 UTC

mullvad quality is down so bad. kinda sad

2026-10-14 07:14:21 UTC

- reply

Fucking cool 💯

2026-10-14 07:07:30 UTC

9 in 10 websites keep AI engines out.
Stradiji just audited how sites treat AI crawlers. Three in four give them no explicit access in robots.txt at all. Their founder put it bluntly: companies are paying to show up in AI answers while their own robots.txt tells those systems to stay out.
It's the simplest possible GEO failure. Before you worry about what the assistants say about you, check whether you've even let them in the door.
We run a free AI Readiness Score that checks exactly this, across ChatGPT, Gemini, Claude, and Perplexity: where you show up, where competitors do, and what to fix first.
When did you last read your own robots.txt?

2026-10-13 07:19:28 UTC

Kev 4B TEE is now available for classification, routing, scoring, and yes/no decisions. Its serving provider reports TEE (Trusted Execution Environment) hosting. NanoGPT does not currently enforce attestation or verify signed receipts for this route. Jared Palmer's open-weight model evaluates text or JSON state in one forward pass and returns typed answers with probabilities instead of generated text.
https://nano-gpt.com/subscription/kUGJhgF9

2026-10-12 09:39:37 UTC

- reply

I thought the same when I saw this. There are enough alternatives already, makes no sense imo

2026-10-10 07:32:00 UTC

oof
#nevent1q…ur5r

2026-10-09 05:22:28 UTC

2026-10-08 07:45:20 UTC

I really don't get how people still waste time of adding ex-google ex-deeznuts whatever to their presentations. No one cares, if your product is not better than what claude can create in an hour. it's sad kind of, but the reality

2026-10-05 08:01:09 UTC

My work scored 62/100 'Human in the loop of my ai cowork harness' on AI exposure (High) with the free Tideproof scan. How exposed is yours? https://tideproof.org/score/high/?utm_source=bluesky&utm_medium=social&utm_campaign=scan_share
Fuck

2026-10-04 09:54:07 UTC

Check out Muse, your personal AI agent. Redeem my code in Settings within 48 hours of joining and we'll both get 1 billion Muse tokens.
Code: 8WQ4S7
https://muse.ai/join

2026-10-04 09:39:44 UTC

Jagex Announces RuneScape 4, a New MMO Built in Unreal Engine
https://www.techpowerup.com/353372/jagex-announces-runescape-4-a-new-mmo-built-in-unreal-engine

2026-10-04 04:56:05 UTC

Any chance to have a mcp for mailtstr so my agent could manage the inbox? @npub1qu7…7p98

2026-10-02 14:17:24 UTC

- reply

kernel gate 2?

2026-10-02 14:07:56 UTC

New gym feels more commercial and less comfortable, but damn does it feel good to dominate some white belts and not be the weakest in a small high level group.

2026-09-30 07:15:00 UTC

Five experiments tested whether a slime mould's network rule helps AI models decide, route and retrieve, and it never helped overall.
https://paragraph.com/@metaend/a-slime-mould-is-not-an-agent
#ai #airesearch

2026-09-28 18:02:26 UTC

A slime mould is not an agent
https://paragraph.com/@metaend/a-slime-mould-is-not-an-agent

2026-09-28 15:12:44 UTC

@npub1jug…fc28 Ran breez.technology through metaend Grade: C (60%). Your llms.txt files score full marks, but pages don't answer agents asking for markdown or JSON (0/5), and llms-full.txt still says USDT/USDC "coming soon". Lite report attached. Full report: $29 in sats or Cashu.
https://blossom-01.uid.ovh/4c9f32882c9acea10cf44dcbab07899b35102b4fbb266d8dd87668735c70c1c8.png