Hi, I'm Tomas Hajek.

Projects

Featured

• The token tax of Norwegian Bokmål in LLMs

  Published: 5 May, 2026

  • 2 min read

  The same government foreword used 37.4% to 67.5% more tokens in Norwegian Bokmål than in English, despite fewer characters.

• AI agents are useful for old software backlogs

  Published: 27 Apr, 2026

  • 2 min read

  Agents can inspect stale bugs, duplicate issues, old PRs, and half-finished ideas without getting bored. The decision still needs evidence.

• Fixing OpenClaw with OpenClaw

  Published: 9 Mar, 2026

  • 2 min read

  OpenClaw helped investigate and fix its own bug. The agent shortened the path to a reviewed patch, while humans kept control.

• What years of work with Gjensidige taught me about customer communication

  Published: 12 Dec, 2025

  • 4 min read

  What long-term CCM work with Gjensidige taught me about technical memory, trust, accessibility, and controlled communication.

• DocEngine now drafts accessibility templates with AI

  Published: 28 Aug, 2026

  • 2 min read

  Upload a sample PDF, get a drafted accessibility template. Available today in DocEngine, with two non-negotiables built in.

• The model names things, the code keeps the coordinates

  Published: 24 Aug, 2026

  • 4 min read

  We tried drafting accessibility templates without an LLM first. What shipped gives the model meaning but never a coordinate.

• The config line that outlived its model

  Published: 23 Aug, 2026

  • 2 min read

  I burned 39% of a weekly Codex quota in four and a half hours. A context-window override I set during GPT-5.5 was still in my config.

• What a 14-hour Codex run taught me about task queues

  Published: 20 Jun, 2026

  • 4 min read

  After a 14-hour Codex run, I want long agent tasks controlled by a queue with a clear finish line and reviewable evidence.

• The review bottleneck is the new software bottleneck

  Published: 18 Jun, 2026

  • 7 min read

  AI agents increase code output faster than teams increase review capacity. Senior attention moves to understanding, verification, and ownership.

• Claude Fable 5 refused my defensive security review

  Published: 11 Jun, 2026

  • 3 min read

  Claude Fable 5 refused a defensive review of our own SaaS application twice, including a prompt that explicitly banned exploit guidance.

• DocEngine now validates PDF/UA with veraPDF

  Published: 2 Jun, 2026

  • 3 min read

  DocEngine now runs veraPDF after processing a document and reports PDF/UA-1 conformance and failed checks in the same workflow.

• Making a 3D-printed lamp stand with my son

  Published: 27 May, 2026

  • 2 min read

  My son wanted a hanging solar bulb on his desk, so we used an AI agent, parametric CAD, and a 3D printer to make a stand.

• Taking ownership of your AI coding history

  Published: 19 May, 2026

  • 2 min read

  I added Windows support to a tool that exports local coding-agent history. The output is useful, but it must be treated as sensitive.

• Warp as an ACP client

  Published: 17 May, 2026

  • 3 min read

  I built a prototype of Warp as an ACP client, with Warp handling the terminal UI and a local process owning the agent runtime.