in10x

in10x / about

Theis Frost

Copenhagen

I build AI agents for real tasks, test them and write up what happened. in10x is where I show the work and explain how I did it.

how I work

  • Agents do the building. AI agents write most of the code and collect most of the data. I set the task, review the result and decide what ships.
  • One small server. Everything runs on a server I look after myself, with plain static sites where that is enough.
  • Approval first. Agents draft. Nothing sends a message or spends money without a person approving it.

what I like building

Systems that learn from their own results. Each run leaves notes for the next one, and a person approves what becomes a rule. The front page has four examples.

the work

  • AI citation research. Which sources AI assistants cite, by market and model. Pipeline, study and dashboard at aigeoaudit.com.
  • Fontvera. EU digital law as structured data, with search and an MCP server.
  • Fontfore. UK procurement data, collected daily, with a Monday brief.
  • Ladeagent. A customer service agent where the model is chosen by evals.
  • Agentis. Danish meeting minutes from audio, with an approval step.
  • One Study a Day. Research papers turned into fact-checked videos and study pages.
  • Dial Dynamo. A voice agent for home-service trades.
  • The lab. Every experiment since December 2025, including the ones that stopped.
  • The playbook. Posts on how I build and test agents, with these projects as the examples.

get in touch