About

Tomas Korcak (korczis), software engineer in Brno: what I build, the languages I build it in over the years, and where to find it.

Tomas Korcak

korczis

Software engineer · Brno, Czech Republic

On GitHub since: August 2010
Original public repositories: 211
Contributions, 12 months: 20,356
Most used language: JavaScript (30)
Author of: Majordomus

GitHub GitLab majordomus.dev

What I build

For 16 years, in public: tools that make AI agents prove their work, honest scientific instruments in the browser, knowledge bases you can check, data catalogues and maps, and the small tools in between. The counts are every original repository on GitHub and GitLab, by inferred category.

AI engineering7 Supervision for AI coding agents: one policy for every tool, scoped tasks, and work that lands only when it is proven. Data and GIS47 Catalogues and maps of open data: geodata sources, incident reports geocoded and served from one registry. Knowledge8 Knowledge bases you can check: cross-linked taxonomies with their sources, built as books, sites and graphs. Learning and experiments16 Learning materials, katas and experiments across many languages: the sketches that came before the projects. Libraries10 Reusable building blocks for other programs: clients, bindings, validators, parsers and framework extensions, mostly Elixir and Rust. Media and art8 Work made from code: a promo reel rendered end to end, a music blog, and a digital artifact or two. Science and visualisation14 Honest instruments and drawn algorithms in the browser: acoustics, physics, sorting, mazes, each readout measured. Security and OSINT4 Open-source intelligence, radio and network security research: cases, proofs of concept and the tooling around them. Systems and infrastructure16 Servers, clusters and the plumbing beneath software: immutable hosts, packet processing, embedded boards and deployment. Tools72 Small, sharp command-line tools and indexes: data conversion, CSV to SQL, gist search. Web9 Websites and web applications: single-page apps, Phoenix and Elm frontends, landing pages and the sites of friends.

Languages over the years

Original repositories by year of creation, as streams of their languages; a stream opens its repositories.

Show as a list

Languages, as bubbles

Every original repository by its primary language; a bubble opens its repositories.

Show as a list

Skills

Derived, not claimed: every skill links to the work that shows it. Hover or focus a skill to see its evidence.

Languages

Technologies

Domains

Selected work

All projects

Majordomus · 2026 · Rust

AI agents say the work is done; Majordomus makes them prove it.

Oscilla · 2026 · JavaScript

A zero-install acoustic laboratory in the browser.

algoglyph · 2026 · TypeScript

Algorithms and physics drawn as glyphs.

Kompendium · 2026 · Elixir

A defensive taxonomy of epistemic and cognitive manipulation (Czech).

Data Atlas · 2026 · Python

A searchable catalogue of GIS and geodata sources (Czech).

Majordomus reel · 2026 · HTML

A promo reel rendered entirely from code.

Sources

Profile, contributions and languages: GitHub API, 14 November 2026.
Repositories and authorship: GitHub and GitLab APIs through gh and glab, 14 November 2026; original = the owner wrote at least half of the listed commits; forks and mirrors are counted, not listed.
Majordomus release: majordomus.dev/build.json, v1.0.2.
Method and every derivation: Stack, data and the source of this site.

Article

A working method

The repositories tell the story better than a biography could. The language streams on this page start in Ruby and JavaScript, the tools of the early web, widen into Elixir for systems that must keep running, and arrive at Rust where correctness and speed both matter. The constant is the method, not the language.

That method is evidence over claims. In OSINT work an answer without a source is worthless; the same rule runs through the software here. A number on a page is measured, a page that follows from data is generated, and a claim that something works is backed by the check that proved it.

The newest work applies that rule to AI agents. Majordomus gives every coding agent the same policy, scopes each task, and lets work land only when its verification ran; this site is built under it. Most of the code is open source, so every statement here can be followed to the repository that shows it.

Concepts on this page

The terms this page builds on, each defined in the glossary with a source to read further. All terms

MajordomusTechnology: The owner's supervisory layer for AI-assisted work: one policy for every coding agent, scoped tasks, and finishes that must be proven.
AI agentConcept: A program that uses a language model to plan and act (edit code, run commands) toward a goal; Majordomus makes such agents prove their work.
OSINTConcept: Open-source intelligence: knowledge built from publicly available sources, with every claim traceable to its source.
Open sourceConcept: Software whose source code anyone may read, use, change and share under its licence.
RustProgramming language: A systems language that guarantees memory safety without a garbage collector; Majordomus is written in it.
ElixirProgramming language: A functional language on the Erlang virtual machine, built for concurrent, fault-tolerant systems.
RubyProgramming language: A dynamic, expressive language designed for programmer happiness; many of the older tools here are Ruby.
JavaScriptProgramming language: The programming language of the web browser, also run on servers through Node.js. Every interactive part of this site is JavaScript.