---
description: Short posts about things that caught my eye: AI, technology, reading tips
url: https://www.schmalzl.tirol/en/blog
lang: en
deutsch: /blog.md
---

# Blog

[To the notes from the old blog](#notizen)

Short posts whenever something catches my eye: an article worth reading, a number that sticks, a tool that saves work. Newest entries first. Use the tags above the list to filter, and the control next to them to sort.

Archive posts carry the date they were originally found. A note at the end mentions that they were written up on 11 October 2026. The topics range from AI and knowledge collections to databases and web design, and on to 3D, robotics and retro technology.

## Pages

- [22 years of Amazon – what my orders say about me](/en/blog/2026-10-18-amazon-seit-2005.md): 18 October 2026 · Review
- [Building games and 3D worlds with AI](/en/blog/2026-10-15-spiele-und-3d-bauen-mit-ki.md): 15 October 2026 · Thoughts
- [Every app is now open source, whether it likes it or not?](/en/blog/2026-10-15-jede-app-ist-jetzt-open-source.md): 15 October 2026 · Thoughts
- [Finds #1: animation at a click, documents as Markdown and a life in weeks](/en/blog/2026-10-15-fundstuecke-1.md): 15 October 2026 · Find
- [More AI, less free time?](/en/blog/2026-10-15-mehr-ki-weniger-freizeit.md): 15 October 2026 · Thoughts
- [One person, many agents](/en/blog/2026-10-15-ein-mensch-viele-agenten.md): 15 October 2026 · Thoughts
- [Humanoid robots with character](/en/blog/2026-10-11-humanoide-roboter-mit-charakter.md): Steve, Miroki, Miroka and Reachy show how likeable and quirky robots can look.
- [My AI editor, and I am the boss](/en/blog/2026-10-11-mein-ki-redakteur.md): 11 October 2026 · Workshop
- [OpenCNC3D: the Nintendo 64 Command & Conquer comes to the PC](/en/blog/2026-10-11-opencnc3d.md): 11 October 2026 · Find
- [A website in two evenings and one day](/en/blog/2026-10-09-eine-website-in-neun-stunden.md): 9 October 2026, addenda 11 and 13 October 2026 · Workshop
- [How many people really use AI?](/en/blog/2026-10-09-wie-viele-menschen-nutzen-ai.md): 9 October 2026 · Reading tip
- [What I told Claude Code and Codex while building this site](/en/blog/2026-10-09-meine-prompts.md): 9 October 2026, addenda 11 and 13 October 2026 · Workshop
- [Who holds Tyrol's AI scene together](/en/blog/2026-10-08-ki-koepfe-in-tirol.md): The people behind KINN, Die KI Woche and the meetups in the Unterland
- [A playful 3D room as an interface](/en/blog/2026-10-07-spielraum-3d-oberflaeche.md): A spatial view can show relationships, but it needs unambiguous states and simple controls.
- [A detection demo with invented parcels](/en/blog/2026-09-24-synthetische-pakete-erkennungsdemo.md): Synthetic scenes help to explain things, but they do not replace testing under real conditions.
- [KIDU: 25 prototypes in one trade fair day](/en/blog/2026-09-22-kidu-buildathon-fafga.md): 22 September 2026 · Event
- [SEO vs. GEO: being found when AI gives the answer](/en/blog/2026-09-05-seo-vs-geo-thomas-winkler.md): 5 September 2026 · Reading tip
- [Die KI Woche: news and podcast](/en/blog/2026-08-20-die-ki-woche.md): 20 August 2026 · Reading tip
- [KIKOM and Markus Kirchmair: AI consulting from Tyrol](/en/blog/2026-07-15-kikom-markus-kirchmair.md): 15 July 2026 · Reading tip
- [Turning your own test page into structured data](/en/blog/2026-06-14-testseite-strukturiert-extrahieren.md): Selecting titles, text and table values deliberately and keeping their origin.
- [Operating a sample page with a browser agent, under control](/en/blog/2026-06-10-browseragent-eigene-testseite.md): Your own test page makes the goals, the permitted actions and the visible completion of a browser task verifiable.
- [Bringing your own drawing to life](/en/blog/2026-06-04-eigene-zeichnung-animation.md): Movement is more convincing when the start, the goal and the shape of the drawing fit together.
- [KINN: Thursdays at 8, Tyrol's AI community](/en/blog/2026-06-04-kinn-tirols-ki-community.md): 4 June 2026 · Community
- [Asking a table and checking the answer](/en/blog/2026-05-20-tabelle-fragen-antwort-pruefen.md): Questions in natural language only become verifiable through traceable data and queries.
- [Supermarket discounts only with an app: a question of clarity](/en/blog/2026-05-14-rabatte-app-alltagsfrage.md): Price conditions, sign-up and extra effort should be clear before you decide.
- [A home server of your own: the task first, then the software](/en/blog/2026-05-09-heimserver-aufgabe-software.md): A manageable purpose helps to make sense of containers, virtual machines and self-hosted services.
- [A small toolkit for JSON and text patterns](/en/blog/2026-05-08-json-regulaere-ausdruecke-werkzeugkoffer.md): Formatting, validating and searching precisely: three different tasks that are easy to mix up.
- [Open-weight AI models and the question of choice](/en/blog/2026-05-06-offene-modelle-wahlmoeglichkeiten.md): Availability, licence, modifiability and actual operation are different properties.
- [A PDF is a view; data needs a structure](/en/blog/2026-04-23-pdf-text-struktur-lesereihenfolge.md): Why text recognition, reading order and table structure should be checked separately.
- [A small diagnostic tool needs a specific question](/en/blog/2026-04-23-terminal-werkzeuge-konkrete-frage.md): How to get to know a tool without getting lost in a large collection of features.
- [Finding public sources and recording where they come from](/en/blog/2026-04-20-oeffentliche-quellen-herkunft.md): A source link only becomes useful with context, a date and a clearly assigned statement.
- [A whole working environment grows around a language model](/en/blog/2026-04-19-ki-plattform-modell-werkzeuge.md): Interface, model connection, data and tools each do a different job.
- [Local AI: model, memory and task belong together](/en/blog/2026-04-19-lokale-ki-modell-speicher-aufgabe.md): Why the right local AI depends on the specific task and the whole setup.
- [Understanding AI usage without pricing promises](/en/blog/2026-04-13-ki-verbrauch-umfang-parallelitaet.md): Small single tasks and large repeated workflows can create very different amounts of work.
- [From sketch to 3D idea: what does a demo really show?](/en/blog/2026-03-22-skizze-3d-modell-demo.md): A convincing image, an editable model and a design ready for manufacturing are different results.
- [Small icons, big differences](/en/blog/2026-02-05-icons-konsistent-gestalten.md): Why a symbol should not just look nice, but be recognisable in every state.
- [Electric mobility: the task belongs in the comparison](/en/blog/2025-10-24-elektromobilitaet-energie-aufgabe.md): Drive, energy storage and actual use are different parts of the same question.
- [Playful interfaces for running AI tasks](/en/blog/2025-10-19-spielerische-oberflaechen-ki-aufgaben.md): Animations can make states visible as long as they keep a recognisable meaning.
- [What an AI agent should remember](/en/blog/2025-10-18-ki-agent-gedaechtnis-skills.md): Providing task rules and project knowledge selectively instead of storing everything indiscriminately.
- [Assistant, workflow or agent: who decides the next step?](/en/blog/2025-10-13-assistent-workflow-agent.md): Three ways of working, explained with an invented packing list for a day trip.
- [From AI prototype to a reliable small application](/en/blog/2025-10-04-ki-prototyp-verlaessliche-anwendung.md): Why a working demo is only the beginning: goals, error cases and traceable checking steps.
- [A portable knowledge file for agents](/en/blog/2025-09-26-portable-wissensdatei-agenten.md): The idea of a single knowledge file raises practical questions about origin, updates and findability.
- [Looking at AI progress in verifiable steps](/en/blog/2025-09-07-ki-fortschritt-ueberpruefbare-schritte.md): There is an important distinction between a forecast and a task that has actually been solved.
- [How AI texts become easier to understand](/en/blog/2025-08-06-ki-texte-verstaendlich-ueberarbeiten.md): A style sample of your own, a clear reader and concrete statements instead of interchangeable phrases.
- [One keyboard for several computers: the transitions count](/en/blog/2025-07-19-tastatur-mehrere-computer.md): Shared input is only pleasant if the target, focus and limits stay clear.
- [Speaking instead of typing when drafting](/en/blog/2025-07-13-sprechen-text-aufgabe.md): A sample text I dictated myself shows where dictation helps and why the result has to be proofread.
- [Why automations have to stay maintainable](/en/blog/2025-06-15-automatisierungen-wartbar-halten.md): A process needs understandable triggers, error states and a clear end point.
- [Clear goals make tasks and meetings easier to follow](/en/blog/2025-06-13-organisation-klare-ziele-aufgaben.md): An invented community project shows why time, decisions and responsibilities belong together.
- [A digital ticket needs the right details above all](/en/blog/2025-06-10-digitales-ticket-wichtige-angaben.md): An invented ticket shows how data structure and readable presentation work together.
- [Notes as files: a small knowledge collection](/en/blog/2025-06-04-notizen-markdown-wissenssammlung.md): How links, traceable sources and ordinary text files turn notes into a usable collection.
- [Why robots do not have to look human](/en/blog/2025-05-30-robotik-aufgabe-statt-menschenform.md): The task decides a robot's shape, movement and how it works with others.
- [Learning with AI: one small task instead of ten courses](/en/blog/2025-05-21-ki-lernen-kleine-aufgabe.md): A checkable learning goal makes getting started clearer than a large collection of courses and tools.
- [Reading new software as an archive find](/en/blog/2025-05-09-software-archivfund-einordnen.md): Why the date of the find, the announcement at the time and today's product are not the same thing.
- [A good AI task needs a goal, context and examples](/en/blog/2025-05-04-ki-aufgaben-klar-formulieren.md): A concrete brief helps more than a long collection of supposed magic words.
- [Retro tech: what hides in documents and old game code](/en/blog/2025-01-25-retro-dokumente-spielcode.md): Technical experiments and public source code show other sides of familiar formats.
- [FileMaker and JSON: keeping data and changes readable](/en/blog/2024-10-17-filemaker-json-versionen-dokumentation.md): Structured sample data, traceable changes and small interfaces instead of opaque exports.
- [Spraying or rolling paint: understand the task first](/en/blog/2024-09-04-renovierung-werkzeug-arbeitsweise.md): A different way of working also changes the preparation, the masking and the clean-up.
- [Why a preview image is no substitute for checking the content](/en/blog/2024-08-13-vorschaubild-keine-inhaltspruefung.md): A short description can show a topic, but it cannot prove every claim or an unnamed tool.
- [Spatial displays: a first impression is not an everyday test](/en/blog/2024-02-10-raeumliche-anzeigen-demo-alltag.md): A product demo, a prototype and your own use answer different questions.
- [A message as HTML: structure before decoration](/en/blog/2023-02-22-html-nachrichten-struktur.md): A readable message needs clear sections and meaningful links, not design stunts.
- [Links, lists and charts in a small database](/en/blog/2023-01-10-filemaker-links-listen-diagramme.md): An invented book list shows how display and data have to fit together.
- [How a web interface keeps its style](/en/blog/2022-07-21-weboberflaeche-designsystem-zustaende.md): Record colours and spacing, but also design keyboard use, errors and empty states.