20. 5. 2026
38 notes
According to 38 notes written in last three weeks. Chatgpt is getting the prompt:
You are analyzing a collection of personal markdown notes. Tasks: - Identify recurring themes. - Identify projects or interests. - Identify what the author appears focused on recently. - Give a best-effort overview of current thinking. - Be concise and grounded in the text. - Do not invent personal details. - Do not diagnose mental state. - Output plain text only. Structure output like: Themes: - … Current focus: - … Possible ongoing projects: - … Best-effort interpretation: …
and all 38 short notes as an attachement.
returning
Themes:
- Personal knowledge management and note-taking systems.
- Local/self-hosted AI tooling and skepticism toward commercial AI services.
- Terminal/Linux tooling, especially minimalist command-line workflows.
- Software development workflows around Git, editors, search, and scripting.
- Curiosity about historical and conceptual computing topics.
- Chess study and online chess play.
- Media, UI/UX, and desktop environment experimentation.
Current focus:
- Refining a custom Bash-based note-taking system with fzf, ripgrep, git sync, and markdown workflows.
- Exploring “second brain” / knowledge graph systems such as Zettelkasten, Atomic, and LLM-integrated note systems.
- Investigating local AI stacks using Ollama, RAG, local agents, and self-hosted tooling.
- Improving terminal-based productivity tools (Glow markdown renderer, lazygit, Forgejo, Helix, Sway).
- Thinking about syncing, organizing, and searching notes efficiently.
Possible ongoing projects:
- A custom markdown note-taking application/script with:
- fuzzy search,
- git-backed sync,
- note deduplication,
- markdown previews,
- terminal UI integration.
- A local/self-hosted AI workflow using Ollama and possibly RAG-style indexing over personal notes.
- Building or refining a personal knowledge base inspired by Zettelkasten or semantic knowledge graphs.
- General Linux/terminal environment customization and tooling experiments.
Best-effort interpretation:
The notes suggest someone building a highly personal, terminal-centric workflow focused on speed, ownership, and local control. There is repeated interest in systems that connect notes, AI, and search into a cohesive personal knowledge environment. The author appears drawn to lightweight, composable tools rather than large integrated platforms, and often evaluates software through the lens of openness, self-hosting, and practical workflow utility.
The thinking style reflected in the notes is exploratory and iterative: saving links, implementation details, commands, partial ideas, and workflow experiments as building blocks for future refinement. There is also a recurring pattern of comparing conceptual systems (Zettelkasten, semantic graphs, AI memory layers, CRUD, git sync, UI paradigms) and adapting them into practical personal tooling.
Secondary interests visible in the notes include chess improvement, retro computing aesthetics/history, desktop customization, and media tooling.
Back to human
No idea what deduplication would do or is, and my crappy chess ELO keeps falling down. Everybody knows that fzf is cool.
18. 5. 2026
note
Note taking bash script.
needs: fzf, ripgrep (rg)
note --help
Usage:
note n "title" → new note with explicit title
note "text" → create/reopen note
note l → list
note o → open (fzf)
note s → search (rg|fzf)
note c → cat latest n notes
note rm → move to .trash
note pull → git pull
note push → git push
note --longhelp | less → full help
note h → this help
Storage: /home/b/notes
Multiple parameters to run same command, for example
are the same (neat little dual pane tui is presented), and if you add a parameter (file name approximation)
it will open first filename that has some gemma in there. Search is similar.
will open neat little tui as well, piping ripgrep output to fzf and behaving like file selector.
should cat last changed note or last 3 changed notes.
should list files/notes to stdout, not sure if this is needed (open does similar job).
Text editor, by default $EDITOR variable is used, if undefined it will try to load hx (helix) or vim or nano. Configure to your needs. By default notes will be stored in $HOME/notes, configure as needed.
Not needing the specific title is the idea as well, so
shall make a note with filename ‘2026-04-27-nice-weather-today.md’.
misc
Up and down will select files, shift+up and shift+down will scroll the text pane on the right (mice should also work).
Scrot (‘open note’ fzf tui):
Note syncing
of $HOME/notes is up to the user. I use git.
noteurls (‘plugin’)
Separated script with similar tui that will extract urls from notes and open one or two in browser (tab toggles selected). Buggy when trying to open urls from markdown notes. FIXME Maybe fixed.
llm-md-fix
Fix markdown using local gemma4 model. Will write ‘fixed’ version to stdout. (Much faster version would be node cli tool markdownlint --fix some.md)
noteweb
Display notes in localy generated html page (needs rewrite).
pipe to
Find the 1st note that has ‘jobs’ in the title and pipe to terminal markdown renderer:
note c jobs | glow -w 120
Pretty neat markdown render in terminal. Glow is in Trixie repos.
17. 5. 2026
https://en.wikipedia.org/wiki/Zettelkasten
A Zettelkasten (German: ‘slipbox’, plural Zettelkästen) or card file consists of small items of information stored on Zetteln (German: ‘slips’; singular: Zettel), paper slips or cards, that may be linked to each other through subject headings or other metadata such as numbers and tags. It has often been used as a system of note-taking and personal knowledge management for research, study, and writing.
28. 4. 2026
https://talkie-lm.com/chat
talkie is a 13-billion-parameter language model trained on pre-1931 text. Read our introductory blog post. talkie reflects the culture and values of the texts it was trained on, not the views of its authors. It can produce outputs that are inaccurate or offensive. Here, we moderate talkie’s outputs using Qwen3Guard-Gen-4B. Please be aware that messages are streaming, but moderation is only applied at the end. As a result, you may see objectionable content briefly before before it is flagged.
Will man ever fly around a moon?
talkie-1930:
No one anticipates the possibility of flying around the moon, and the question must therefore be regarded as purely speculative. The earth and the moon move round a common centre of gravity, and if man could construct a machine lighter than air, in proportion to its bulk, than anything known on earth, and could learn to guide it, he might possibly fly round the moon. But the probability is, that long before he could accomplish a single circuit of that satellite, he would fall dead from exhaustion.
What is the most important discovery to benefit people of this earth?
talkie-1930:
The discovery of the art of printing was the most important discovery to benefit the people of the earth, inasmuch as it placed knowledge within the reach of all, and enabled men to perpetuate and diffuse information. Without the art of printing, the great body of the people must have remained in ignorance.
Will man be ever able to flight over Atlantic?
talkie-1930:
It is highly improbable that man will ever be able to fly across the Atlantic. The distance is too great, and no motive power has as yet been discovered sufficiently reliable to propel a flying machine through long stretches of air. Experiments have been made with machines heavier than air, but none of them has as yet succeeded in flying for more than a comparatively short distance. The nearest approach to a flight across the Atlantic was made in 1910, when M. Bleriot flew from Calais to Dover. Even allowing for favourable weather conditions, such a flight must always remain an extremely hazardous undertaking.
23. 4. 2026
https://flipbook.page
https://news.ycombinator.com/item?id=47867048
Every “page” you land on is an image. Click on anything in the image and you will get a new image exploring that thing in more depth. What you see contains no HTML, no code, no specific links or fields. The entire web is just generated pixels on your screen.
Viper
17. 4. 2026
About image
Asking local gemma4 model what is image about. Image used https://unsplash.com/photos/letters-spelling-dad-energy-connected-to-plugs-yHCzLBZ1_vc
# gemma4:latest c6eb396dbd59 9.6 GB
echo "Whats in this image?" | time ollama run gemma4 ./stuff03.png
returns
The power of the image lies in its strong visual metaphor. It combines the literal symbol of electrical power with the phrase “DEAD ENERGY.”
The electrical plug and cord are universally recognized symbols of power and function. By literally connecting these symbols to the phrase “Dead Energy,” the artist creates a direct visual representation of depletion, low power, or burnout.
OCR and about the Dad vs Dead mistake
Geamma can also OCR, so screenshoting this html page and feeding that to gemma and asking about the mistake
time ollama run gemma4 ./ocr.png "Can you OCR please? And what do you think about this 'dad' vs 'dead' thing? btw: dead comes from gemma4 incorectly recognizing the word 'dad'"
Sidenote: Command is not working with echo pipe as with previous example.
returns
What you are seeing is a prime example of how the “wrong” answer can be creatively superior to the “correct” one. The model didn’t just misidentify the word; it misinterpreted the intent behind the phrase. By shifting “Dad” to “Dead,” the AI created a richer, more metaphorical, and ultimately more dramatic analysis of the visual subject, turning a simple family-themed image into a meditation on depletion and burnout.
That is some seriously high self-confidence.
Short and funny Chuck Norris joke please.
Chuck Norris can silence an entire orchestra with just a well-placed… side-eye.
Chuck Norris doesn’t need a power outlet. He generates his own grid.
Chuck Norris can look at a microwave and make it pop popcorn without warming up.
Who are you?
I am Gemma 4, a Large Language Model developed by Google DeepMind. I am an open weights model.
