Seneca wrote two thousand years ago about attention, time, discipline, and voluntary hardship. He was describing your AI deployment. By Enzo Duit.
Seneca's warning about dispersed attention applies with startling precision to the founder who deploys AI agents for everything at once. I made this mistake. The agents were fine. My specifications weren't.
While the AI world announces three new models a week, the founders who build durable systems are doing something unfashionable: ignoring most of it. Seneca called this discipline two millennia ago.
Everyone has an opinion about AI replacing jobs. I stopped listening. Running three companies with AI agents as my team isn't a hot take — it's a Tuesday. The crowd was not helpful.
Seneca's first letter opens with a demand: account for your time, defend it, recover it. $120 a month on AI infrastructure that returns hundreds of hours of human attention is not a cost. It's an arbitrage.
At kilometer 65 of the Ushuaia 130K, my knee gave out. I ran to kilometer 90 on painkillers. The race was eventually suspended there — but I didn't know that yet when I chose not to quit. Neither does your agent.
You do not control what an AI agent outputs. You control the specification. This is not a limitation — it is the entire discipline. Seneca understood this about life. It applies to production deployments with uncanny precision.
Enzo Duit is an Austrian entrepreneur based in Buenos Aires. He runs multiple companies — Trillion Initiative, Fly Raising, Agent School — using AI agents instead of employees, at roughly $120/month total infrastructure cost.
His frameworks: OFA (Output-First Architecture) and FOA (Founder on AI) reflect his conviction that the failure mode in AI deployment is almost always the specification, not the model.
He is also an ultra runner, preparing for Val d'Aran 110K in July 2026.
Website: enzoduit.com