Michael Christen / YaCy Search

Open Source Developer and Search Engine Creator,
Maintainer of YaCy.net and SUSI.ai, Research on AI in Information Retrieval
Follow me on Patreon for news about YaCy: https://www.patreon.com/orbiterlab

github: https://github.com/orbiter
patreon: https://www.patreon.com/orbiterlab
YaCy: https://yacy.net

Public Key

npub1v5nnv7827xa34r5fmvppkq69vpgjm27cva4dle3s53st2gg8cetqkeldnk

NIP-05 Address

orbiterlab_at_sigmoid.social@momostr.pink

Profile Code

nprofile1qqsx2fek0r40rwc636yakqsmqdzkq5fd40vxw6klucc2gc94yyruv4spz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dsvawssn

Publishing to

relay.damus.io

nos.lol

Author Public Key

npub1v5nnv7827xa34r5fmvppkq69vpgjm27cva4dle3s53st2gg8cetqkeldnk

Show more details

Last Notes

2026-10-18 12:28:50 UTC

gpt-6.1-sol-think saturates PE-Bench-200-Java from https://github.com/Orbiter/project-euler-llm-benchmark : Python 99%, JavaScript 99.5%, Java 100%, Rust 99%, Clojure 87%. The first model in this benchmark to solve all of Project Euler’s first 200 problems from https://projecteuler.net/

2026-09-25 14:39:48 UTC

vcoding useless? game 4 YaCy...
https://cdn.masto.host/sigmoidsocial/media_attachments/files/117/332/188/435/247/886/original/0f44279c4b5340e3.png

2026-09-22 16:54:07 UTC

mflux-server #image #generation for #mac with support for qwen-image-2.1 using mflux 0.20.0 https://github.com/Orbiter/mflux-server
Start the server with
./run.sh --model qwen-image-2.1
Example below uses image editing of the #YaCy logo
https://cdn.masto.host/sigmoidsocial/media_attachments/files/117/315/731/279/944/505/original/28dbb1ab7cd93692.png

2026-09-20 11:50:00 UTC

mflux-server 0.19.2 https://github.com/Orbiter/mflux-server #imagegeneration #mflux #server
https://cdn.masto.host/sigmoidsocial/media_attachments/files/117/303/212/330/033/064/original/130cb2a6fcb01bb5.png

2026-09-02 17:42:54 UTC

The new YaCyStats!
https://new.yacystats.de/
https://cdn.masto.host/sigmoidsocial/media_attachments/files/117/202/678/237/796/396/original/466c0dc417542092.png

2026-08-20 19:58:18 UTC

- reply

ich habe meinen Mitschnitt hochgeladen https://sigmoid.social/@orbiterlab/117129599447071900
#note1wtg…a05v

2026-08-20 19:57:30 UTC

FrOSCon 2026: Baue deine eigenen KI-Agenten mit lokalen LLMs
Mitschnitt vom Workshop zu den KI-Agenten-Tools aus https://github.com/Orbiter/cognitive-utilities
https://www.youtube.com/watch?v=oJ2JXCQTRUk
Wenn ihr wissen wollt wie man mit kleinen Scripten LLMs abfragt, Ergebnisse programmatisch auswerten kann, Linux-typische kognitive Filter baut und auch 1-Script Coding Agents schreibt und eine Stunde Zeit übrig habt...
https://cdn.masto.host/sigmoidsocial/media_attachments/files/117/129/563/989/171/828/original/7e23b3bc4db13b7e.png

2026-08-02 20:59:49 UTC

- reply

stl looks great
https://cdn.masto.host/sigmoidsocial/media_attachments/files/117/027/919/115/218/681/original/61e53df48e564619.png

2026-08-02 18:06:01 UTC

Do you know SketchForge? This is a free/open Tinkercad clone that runs on your own computer. It has a MCP server inside, I attached it to codex and prompted "create a house with garden and a car in front". This is amazing! https://sketchforge3d.com/
https://cdn.masto.host/sigmoidsocial/media_attachments/files/117/027/237/614/125/328/original/55136625fdaf9d8c.png

2026-07-14 07:00:46 UTC

- reply

without moving to Jetty 12, we are stuck with Solr 9. So it has to be done.

2026-07-12 15:27:26 UTC

This weekend I did major work deep down in the basement of YaCy: I replaced Jetty 9 with Jetty 12. Very exciting! Really, this was hard work. And it will enable to move to Solr 10, once we want to use JDK 21.

2026-07-10 17:52:24 UTC

I thought it's time to freshen up the use case images after 20-some years - https://yacy.net gets a new look
https://cdn.masto.host/sigmoidsocial/media_attachments/files/116/896/948/408/083/893/original/a95b81aa96f1c76e.png

2026-07-09 16:46:31 UTC

YaCy has now a complete interface translation for German, French, Spanish, Italian, Greek, Slovak, Ukrainian, Turkish, Russian, Polish, Chinese, Hindi, Japanese, and Korean.
https://github.com/yacy/yacy_search_server/commit/57d188e69a17788ff78fd1b9ad1afa2b232f6536
https://github.com/yacy/yacy_search_server/commit/55956c502b00274edc99888c42a21e2985928fff

2026-06-27 11:33:45 UTC

- reply

I was planning to make Tutorials but I could not find time and my english is not that shiny. Lets see.

2026-06-21 21:27:29 UTC

- reply

yeah, I have a heap of RPi 4 and RPi 3, what should I do with them?

2026-06-21 21:16:56 UTC

- reply

ein RPi 5 16GB, ein RPi4 8GB, zwei RPi 4 4GB. 16 Cores, 32GB Ram. Und ein Mini-Switch.

2026-06-21 20:55:51 UTC

- reply

hatte ich alles rumliegen

2026-06-21 19:15:12 UTC

- reply

preis?

2026-06-21 17:43:16 UTC

Cyberdeck?? Cyberserver!!
https://cdn.masto.host/sigmoidsocial/media_attachments/files/116/789/327/475/956/571/original/8d6b1fc1019480d3.jpg

2026-04-05 20:28:29 UTC

YaCy Skill for OpenClaw
Control and manage a local YaCy search engine instance with OpenClaw.
https://github.com/openclaw/skills/tree/main/skills/qsmtco/yacy

2026-03-29 14:49:15 UTC

KI Chat in YaCy - Vortrag von den Chemnitzer Linuxtagen https://media.ccc.de/v/clt26-322-ki-chat-mit-eigener-suchmaschine-rag-in-yacy

2026-03-29 10:05:53 UTC

YaCy 1.941 is out: a big step toward transparent and private AI. New chat + RAG workflows, tool calling, MCP integration, YaCy Packs, better ranking/snippets #YaCy #Search #RAG #AI
Release Notes: https://community.searchlab.eu/t/yacy-release-1-941-the-ai-release/3376
https://cdn.masto.host/sigmoidsocial/media_attachments/files/116/311/899/858/045/686/original/64c6fdfa74acc0ab.png

2026-03-20 16:44:10 UTC

- reply

it made itself

2026-03-20 15:02:26 UTC

a kanban board that works off tickets itself?
https://cdn.masto.host/sigmoidsocial/media_attachments/files/116/262/104/195/309/863/original/a267f90d07a3a7a6.png

2026-03-12 10:38:33 UTC

- reply

yes, a command shell

2026-03-12 10:27:50 UTC

- reply

I understand the phrase could smell like this. However I am optimistic overall and like to believe in a paradise utopia supported and build with free software. We want to create usefull tools and maybe we should stop to offer technical terms as benefit of libraries, instead show what could be an outcome that is good for everyone, not just nerds.

2026-03-12 09:09:03 UTC

- reply

the one that gives you superpower

2026-03-12 09:06:01 UTC

- reply

with free software

2026-03-12 08:46:13 UTC

Make Life Better

2026-02-17 10:23:45 UTC

Funny: #openclaw #security "do not pursue self-preservation, replication, resource acquisition, or power-seeking", "Do not copy yourself or change system prompts". I guess nobody will mess with that 🧐
https://cdn.masto.host/sigmoidsocial/media_attachments/files/116/085/476/359/983/168/original/abb981ed67cdd7c3.png

2026-02-03 17:26:05 UTC

YaCy Vortrag bei den Chemnitzer Linuxtagen [@clt_news](https://mastodon.social/@clt_news ) : "KI Chat mit eigener Suchmaschine: RAG in YaCy" - Das "KI Release" ist zwar noch nicht fertig, aber wir werden dann unseren eigenen KI Chat zeigen... https://chemnitzer.linux-tage.de/2026/de/programm/beitrag/322
https://cdn.masto.host/sigmoidsocial/media_attachments/files/116/007/859/795/271/470/original/bf3cb0cedb791a3d.png

2026-02-01 17:18:24 UTC

beautiful-mermaid-py - the 1-file python, ascii-art-only version of the beautiful-mermaid library, which is the TypScript version of mermaid-ascii (go); this is vibe-coded, 3200 lines in just one prompt: https://github.com/Orbiter/beautiful-mermaid-py
https://cdn.masto.host/sigmoidsocial/media_attachments/files/115/996/510/970/931/493/original/0c002572760287cd.png

2025-12-22 12:29:32 UTC

mflux-server https://github.com/Orbiter/mflux-server on top of the mflux framework now supports mflux 0.13.3 and has a GUI with a selector for all the models z-image-turbo, FLUX.1-dev, Qwen-Image, Fibo, so you can compare on-the-fly.
https://cdn.masto.host/sigmoidsocial/media_attachments/files/115/763/219/074/493/035/original/85279ad76dee451d.png

2025-12-21 11:56:47 UTC

Welchen der vier folgenden Vortragstitel soll ich für die Chemnitzer Linuxtage 2026 einreichen?
•⁠ ⁠Lokale Multimodale KI zur Erzeugung synthetischer Massendaten: Praktische Werkzeuge für hochwertige Datensätze
•⁠ ⁠KI mit Edge-Computing: Datenhoheit durch lokal betriebene LLMs in mini Cluster
•⁠ ⁠Systemadministration mit KI: Automatisierung von Linux-Tasks mit lokalen LLMs – sichere, eigenständige Workflows
•⁠ ⁠KI mit LoRA: Lokales Fine-Tuning von LLMs für datenschutzkonforme Anwendungen

2025-12-18 21:08:53 UTC

- reply

10€

2025-12-14 09:39:56 UTC

- reply

der m.2 adapter ist billig

2025-12-06 18:52:38 UTC

- reply

I made pre-defined adapters for Ollama, LMStudio, OpenAI and OpenRouter but you can attach anything that speaks the OpenAI API. But... we meanwhile have a full transformer in java in YaCy, it's just not ready yet but will be another option (and its not fast and can only load llama3.2:1b for now). I am also thinking about the option to share your local LLM over the p2p network, so using a "any peer" option might be there in the future.

2025-12-06 18:38:55 UTC

Sneak peek: the upcoming YaCy 1.941 will ship a complete “AI Lab” build environment. Configure every part of your search + LLM + RAG stack from a single UI, keep everything on your machine, zero telemetry. Chat interface included, it's the first AI tool to try out in YaCy. Already working on master. Check out your YaCy dev system and open /AILab.html #SelfHosted #Privacy #OpenSource #AI
https://cdn.masto.host/sigmoidsocial/media_attachments/files/115/674/069/363/449/717/original/d8f002b4494575b7.png

2025-12-05 11:21:29 UTC

Ministral-3 (@MistralAI) performs better than, or on par with, flagship models up to 8 times larger (from just one year ago) on the Project Euler LLM Benchmark (coding):
https://github.com/Orbiter/project-euler-llm-benchmark
ministral-3:14b beats qwen2.5:72b
ministral-3:8b beats llama3.3:70b
ministral-3:3b is on par with gemma3:12b

2025-12-04 14:14:11 UTC

New toy is working! One RTX 3060 is not in a PCIe slot but connected to a M.2 slot via adapter. It runs qwen3-vl:30b-a3b-instruct with ollama and is fast!
https://cdn.masto.host/sigmoidsocial/media_attachments/files/115/661/701/931/904/475/original/6a6f9209351ef237.png

2025-11-25 13:59:00 UTC

Experiments with AI Chat in YaCy
https://cdn.masto.host/sigmoidsocial/media_attachments/files/115/610/689/321/838/487/original/c2c75a229415c67d.png

2025-11-19 13:17:30 UTC

- reply

A lot of technical details from this paper went into the early version of YaCy in 2008, i.e. the Dublin Core Metadata schema, the GET property names for search APIs (using SRU, "Search Retrieval by URL) and the YaCy XML response schema which was called “Opensearch” (but many other things had been named Opensearch later). The json response schema borrows all the concepts from the XML standard, which is simply an extension of RSS.

2025-11-19 13:16:57 UTC

In 2008, I wrote the Machbarkeitsstudie
zum Aufbau und Betrieb einer
„Deutschen Digitalen Bibliothek“ (the German Digital Library)
Here is a thread in the YaCy forum about a request for "Semantic Web" and related technologies.
https://community.searchlab.eu/t/can-or-could-yacy-do-conceptual-indexing-or-be-the-search-engine-for-the-semantic-web/70/26?u=orbiter
It also contains a link to the original paper.

2025-11-18 21:01:36 UTC

Excellent article on running YaCy #DecentralizedSearch #WebScraping in a professional, secure, and scalable way for web scraping and data extraction:
https://scrapingant.com/blog/decentralized-web-scraping-data-extraction-yacy
It covers private/Robinson setups with WireGuard, public-edge reverse proxies (Caddy), container hardening (NIST SP 800-190), crawl profiles, index lifecycle management, performance tuning, and monitoring, JVM/IO tuning, SLOs, ethical scraping, sitemaps-first — everything needed for serious deployments.

2025-11-15 13:44:29 UTC

Open Data in KI nutzen /
Workshop für lokale LLMs mit RAG
Schritt für Schritt die OpenAI-API auf der Kommandozeile kennen lernen.
(repost mit besserem Ton/kürzerem Schnitt von den Chemnitzer Linuxtagen 2025)
https://youtu.be/YU-aUtzrg_Y
https://cdn.masto.host/sigmoidsocial/media_attachments/files/115/553/995/334/682/095/original/1ef3f039d3543440.png

2025-11-15 12:13:16 UTC

- reply

After many years attending the congress I will have family time this year, so I'm not there this year. But a YaCy meet-up is a fantastic idea and you can meet Marc [@m2c_n3e](https://mastodon.xyz/@m2c_n3e ) most probably at the #fossasia table and chat with him about YaCy. I hope that my AI pieces I am working on made it into YaCy until end of december so we can have a chat about it together i.e. with jitsi.

2025-11-09 17:16:10 UTC

- reply

I know that many people reject AI-enriched answers and other services. All of that is optional. You can switch it on if you want to have it, you don't need to do that. YaCy will continue to work without AI.
The "smell" the models have is related to the specific model. You can choose your own. They will improve over time.
Yes, to have the AI functions, you need a (local - good!) provider, which can be ollama or LMStudio. Other options are any OpenAI-compatible APIs.

2025-11-09 13:51:08 UTC

AI Done Right!
YaCy is the idea that we don't surrender control: p2p search built on openness, autonomy, and trust.
Now, as AI becomes the next layer of our digital world, the same principles must hold true.
"AI Done Right" - in YaCy ->
https://www.patreon.com/posts/ai-done-right-143053426
https://cdn.masto.host/sigmoidsocial/media_attachments/files/115/520/060/675/781/691/original/c556658f7e1f532f.png

2025-10-22 20:39:39 UTC

- reply

maybe this is a break in the coding adventure journey that could be filled with "elevator music" or some chiptunes.

2025-10-22 13:25:10 UTC

Can someone make an #AI that composes and plays #music live that gets more and more dramatic while I do #programming and produce code? It should run only locally because it would require to read what I am doing.

2025-10-15 21:43:47 UTC

The Project Euler Benchmark (test how good LLMs perform at code generation) has now a nice graph to show which problems (out of 1..200) can be solved: purple = in all 4 languages (python, java, rust, clojure), blue = in 3, green = in 2, yellow = in 1. https://github.com/Orbiter/project-euler-llm-benchmark/blob/main/README.md
https://cdn.masto.host/sigmoidsocial/media_attachments/files/115/380/362/288/502/150/original/6113bae16c6a5179.png

2025-08-22 00:22:52 UTC

- reply

most functions are now implemented: pack generation, pack download and pack management servlets are ready, a demo pack exists at huggingface and the pack downloader can find and download that pack from huggingface. All datasets at hf with name ' YaCy-Pack' will appear automatically at the downloader servlet. A detailed documentation will be made but is not ready yet.

2025-08-11 23:41:18 UTC

- reply

my impression was that a global index over the fediverse is not wanted

2025-08-02 05:54:43 UTC

We have a YaCy Pack Process Workflow: packs can be added or removed from the search index and therefore are usable both as whitelist and blacklist.
https://cdn.masto.host/sigmoidsocial/media_attachments/files/114/957/618/990/939/518/original/4c9cc9b2fa248f45.png

2025-08-01 17:30:46 UTC

- reply

A first demo pack is now available at https://yacy.net/packs/ and such packs can be imported in YaCy by placing them into DATA/PACKS/load (only with latest YaCy from yesterday). We will provide a more convenient way as well. Pack download locations can be also elsewhere, at huggingface etc.

2025-07-24 04:03:57 UTC

Idea: YaCy data packs - an alternative to web crawling to fill up your search index with pre-packaged well-curated search packages. You could mix and merge those. Everyone could create them. What do you think?
https://cdn.masto.host/sigmoidsocial/media_attachments/files/114/906/221/243/594/752/original/15b1114c3ff38048.jpg

2025-06-29 14:11:43 UTC

Vortrag: "Superhuman AI - Wie und wann ist KI nützlich"
Wie übermenschlich muss Künstliche Intelligenz sein, um unseren Alltag zu verändern? Dieser Vortrag zeigt, dass wir übermenschlichen Super-KI teilweise bereits nutzen können.
https://www.youtube.com/watch?v=sk6AfsF9QlM

2025-06-22 15:33:12 UTC

- reply

every peer published itself every two minutes to a some (say 10, cant remember niw exactly) random other peers. Those return another small number of peers they have seen recently. That distributes peer changes very quickly.

2025-06-22 14:38:07 UTC

New docker release for YaCy - after 8 months (sorry) after a failure of the old release mechanism. Also: new images are now about 400MB instead of 1.3-2GB because of enhanced packaging. And they are multi-arch!
With docker you can run YaCy simply with:
docker run -d -p 8090:8090 yacy/yacy_search_server:latest

2025-06-16 18:06:17 UTC

LLM Model Release Map (red: ChatGPT-3.5-class, blue: GPT-4-class models; guessed). Current progress is about five new major models every month!
https://cdn.masto.host/sigmoidsocial/media_attachments/files/114/694/351/079/980/396/original/6db8f7e7ce238d68.png

2025-04-22 12:18:26 UTC

Wie funktioniert ChatGPT _ganz_genau_? Die verständliche und wirklich vollständige Erklärung in einem Vortrag: https://www.youtube.com/watch?v=T7K2SmqlzOI
https://cdn.masto.host/sigmoidsocial/media_attachments/files/114/381/572/717/672/258/original/ac2f5b391424bb25.jpeg

2025-04-12 16:01:02 UTC

slightly off-topic - I made a thing:
https://www.thingiverse.com/thing:7007980
Pick and Click Bit Rack Reversed
https://cdn.masto.host/sigmoidsocial/media_attachments/files/114/325/824/563/685/463/original/672c14c006464b39.jpg

2025-03-23 14:39:51 UTC

Video vom Vortrag
"Open Data in KI nutzen"
von den Chemnitzer Linuxtagen 2025 [@clt_news](https://mastodon.social/@clt_news ) https://media.ccc.de/v/clt25-269-open-datafreie-daten-in-ki-chatbots-nutzen#t=559
https://cdn.masto.host/sigmoidsocial/media_attachments/files/114/212/258/796/118/905/original/238218171f5bd598.png

2025-03-23 13:56:40 UTC

Folien vom Vortrag
"Superhuman AI - Wie und wann ist KI nützlich"
bei den [@clt_news](https://mastodon.social/@clt_news ) Chemnitzer Linuxtagen 2025 https://yacy.net/material/20250322_CLT2025_Superhuman_AI_Wie_und_wann_ist_KI_nuetzlich.pdf
https://cdn.masto.host/sigmoidsocial/media_attachments/files/114/212/089/123/222/865/original/03ddcd82ad55a4cb.jpg

2025-03-13 01:50:57 UTC

Today, 14:30 (+07) at #FOSSASIA Bangkok:
"The Complete Anatomy of ChatGPT: A Precise Breakdown of LLMs and Transformers"
Tune in live at
https://eventyay.com/e/4c0e0c27/session/9472
Slides can be downloaded from
https://yacy.net/material/20250313_FOSSASIA_2025_The_Complete_Anatomy_of_ChatGPT.pdf

2025-02-22 16:34:20 UTC

- reply

its Solr, it always takes all the given RAM, thats what it does to be efficient.

2025-02-15 11:06:24 UTC

DietPi (lightweight Debian OS for SBCs) comes with #yacy https://dietpi.com/docs/software/distributed_projects/

2025-02-14 15:46:57 UTC

Mit #YaCy auf den #CLT2025 / Chemnitzer Linuxtage - Blog Post von Frank:
https://do3eet.pages.dev/post/clt2025yacy/
https://cdn.masto.host/sigmoidsocial/media_attachments/files/114/003/016/320/294/725/original/4cacd671b76998d8.png

2025-01-09 16:45:50 UTC

New #LLM "PE-Bench" benchmark results for coding, incl. different quantizations and languages: #python, #java, #rust, #clojure using Project Euler challenges: https://github.com/Orbiter/project-euler-llm-benchmark/blob/main/README.md
https://cdn.masto.host/sigmoidsocial/media_attachments/files/113/799/406/055/298/621/original/3f01da3955d770b4.png

2024-12-28 14:52:12 UTC

#DeepSeek-V3 Leading in Superhuman Coding: new leading position in the PE-Bench-Python-100, showing a 15.58-fold #superhuman performance.
See: https://github.com/Orbiter/project-euler-llm-benchmark
https://cdn.masto.host/sigmoidsocial/media_attachments/files/113/731/009/459/900/777/original/f5f54407fef59c0a.png

2024-12-27 14:08:10 UTC

YaCy meetup at #38c3 - I will be at the FOSSASIA table in the Critical Decentralization Cluster, watch out for the YaCy flag! If you want to meet me, just come by. See also: https://community.searchlab.eu/t/yacy-meetup-38c3/1705
https://cdn.masto.host/sigmoidsocial/media_attachments/files/113/725/176/354/633/961/original/4798f5303f369b1a.jpg

2024-12-27 14:01:25 UTC

- reply

You can meet me now or any time in the next four days at 38c3: see my posting in https://community.searchlab.eu/t/yacy-meetup-38c3/1705

2024-12-26 19:12:24 UTC

The PE-Bench-Python-100 test results (checking LLMs ability to code the Project Euler problems) are so far in this table. Read about it in https://github.com/Orbiter/project-euler-llm-benchmark/
https://cdn.masto.host/sigmoidsocial/media_attachments/files/113/720/701/446/097/430/original/0847ca9e8f35d01f.png

2024-12-25 21:47:57 UTC

- reply

ah yes, I just replied this. Your ideas are welcome if you want to contribute.
However my target is to use the test results in a follow-up project where I want to make a auto-coder which reads tickets and provides pull requests. So we can collect here best practices.

2024-12-25 21:44:25 UTC

- reply

the current prompt does not use any COT on purpose because I want to measure also how much different prompt strategies influence the result, so this comes later. Templates are here: https://github.com/Orbiter/project-euler-llm-benchmark/blob/main/templates/template_python.md

2024-12-25 21:33:57 UTC

- reply

my computer is already cooking hard and I want to make a large table. Currently the largest models are first in the queue.
I also want to measure the influence of different quantization levels, because at this time I doubt that usage of a large model with strong quantization is better than a smaller one with less quantization. I don't know any benchmark which shows that.

2024-12-25 21:27:31 UTC

- reply

yes, project euler is around since about 10 years and started with ca. 160 problems. I also believe that most LLMs have seen solutions, however probably mostly in python. I don't think this is too influencing to other programming languages, so benchmarks to java etc. may be more significant.
My target is to evaluate the ability of LLMs to perform in an auto-coder engine and there I would not mind if those models perform because of some 'cheating'.

2024-12-25 19:32:06 UTC

I am creating a LLM benchmark series "PE-Bench-Python-100" to measure how good LLMs are at coding of the Project Euler Challenges compared to humans. The series will include also other languages than python, like java.
The first result so far is interesting because even the small model llama3.2 (3B) has a two-fold super-human performance (score: 2.14)
Source: https://github.com/Orbiter/project-euler-llm-benchmark/

2024-11-29 15:46:47 UTC

- reply

good idea, at #38C3 I am with the #FOSSASIA community and I usually also bring a small YaCy flag to mark where you can find me. We can also do a meetup!

2024-11-26 14:16:01 UTC

only 30 days left until day-0 #38c3 !!1!

2024-11-15 15:37:57 UTC

host your own FLUX.1 server:
https://github.com/Orbiter/mflux-server
This provides image generation with FLUX on a Mac in a multi-user environment. Very easy to set up. Runs fast!
https://cdn.masto.host/sigmoidsocial/media_attachments/files/113/487/706/920/252/558/original/e05b00f83446ede7.png

2024-10-23 14:57:20 UTC

2025 is supposed to become the year of "AI Agents". I collected use cases into a repository that I want to use for lectures. It is still incomplete but it outlines major aspects:
- Inference Building Blocks
- Structured Data Generation
- Agents with internal/external/local data
- Process Automation
- Reasoning and Decision Making
- Robot Steering
- Autonomous Agents
Please have a look:
https://github.com/Orbiter/generative_ai_agents
I would love to get suggestions and maybe help if you want to fill gaps!

2024-10-12 19:22:01 UTC

- reply

voila: https://laufwetter.de/ort/Wipperfuerth/
Nicht dem Link zur DVD-Koordinate im Dokument trauen, der DWD publiziert nur zwei Dezimalstellen der Koordinate der Station.
Happy Running!

2024-10-12 10:14:30 UTC

- reply

Sehr gerne .. leider gibt es in Hückeswagen keinen Messpunkt des DWD, die nächsten sind in Remscheid, Wermelskirchen, Radevormwald und Wipperfürth. Welchen soll ich nehmen? Tolle Landschaft habt ihr da übrigens!

2024-09-10 19:41:00 UTC

I made an API server for FLUX, the new image generator model. There is a very good implementation for Mac Metal, called MFLUX and my server pull request is here: https://github.com/filipstrand/mflux/pull/29
pls give it a thumbs up!

2024-08-19 13:25:00 UTC

Wie funktioniert ChatGPT ganz genau?
Video vom Vortrag bei der @npub1nxl…jpt0 ist nun hier: https://media.ccc.de/v/froscon2024-3054-wie_funktioniert_chatgpt_ganz_genau
https://cdn.masto.host/sigmoidsocial/media_attachments/files/112/988/905/680/072/186/original/01be174dabe0fa43.jpg

2024-08-17 16:39:33 UTC

Wie funktioniert ChatGPT ganz genau?
Die Folien zum Vortrag bei der @npub1nxl…jpt0 sind nun hier: https://yacy.net/material/20240817_Froscon_2024_-_Wie_funktioniert_ChatGPT_ganz_genau.pdf
https://cdn.masto.host/sigmoidsocial/media_attachments/files/112/978/344/194/048/448/original/bb32b8b1b370261c.png

2024-08-16 13:43:27 UTC

Linux macht Spaß uns alles viel besser!
Der Demo-Satz für meinen Vortrag:
"Wie funktioniert ChatGPT ganz genau"
morgen bei der @npub1nxl…jpt0 https://programm.froscon.org/2024/events/3054.html
https://cdn.masto.host/sigmoidsocial/media_attachments/files/112/971/981/535/226/614/original/5b1a6e8ef9b1a9e2.png

2024-08-07 17:43:15 UTC

Mathematik für Neuronale Netze und Transformer ist wirklich einfach, fast primitiv: das allereinfachste aus Linearer Algebra. Man muss nur wissen was Vektoren und Matrizen sind und was das Skalarprodukt ist:
Wir werden das in meinem Vortrag "Wie funktioniert ChatGPT ganz genau?" bei der #FrOSCon @npub1nxl…jpt0 benötigen: https://programm.froscon.org/2024/events/3054.html
Hier ein kleiner Spickzettel. Mehr braucht man nicht um Transformer und damit ChatGPT/LLMs zu verstehen!
https://cdn.masto.host/sigmoidsocial/media_attachments/files/112/921/965/255/809/146/original/f48d304e7cbb2f11.png

2024-08-01 11:32:22 UTC

Save the Date!
FrOSCon 2024-08-17 16:30
@npub1nxl…jpt0
Sankt Augustin (DE)
"Wie funktioniert ChatGPT ganz genau?"
Ich halte mal wieder einen Vortrag und erkläre wie LLMs funktionieren, inklusive (kurzem) Code Review! Alle Open Source Devs sollten das verstehen!
https://programm.froscon.org/2024/events/3054.html
https://cdn.masto.host/sigmoidsocial/media_attachments/files/112/886/536/435/895/574/original/b8b7c487c8b08b32.png

2024-07-31 18:02:29 UTC

Does anybody know how to measure the speed of progress in AI development?
Or a statistic like days per achievement?