← All Letters
Letter V

Per Aspera — The Ushuaia Letter

Inimica est inertiae nobilitas
By Enzo Duit · De Tranquillitate Animi

"Inimica est inertiae nobilitas."

— Seneca, De Tranquillitate Animi — Idleness corrupts the noble mind.

At kilometer 65 of the Ushuaia 130K, my right knee stopped working in the usual sense. Not a cramp. Not fatigue. A structural failure, the kind where the joint simply announces that it is done participating on the terms you have established. I had 65 kilometers left to run. I took painkillers from a checkpoint medic and kept moving.

The race was eventually suspended at kilometer 90 due to conditions on the course. I did not know that at kilometer 65 when I made the decision not to stop. What I knew was that I had chosen to be there. Nobody had put me on that mountain. I had signed up, trained, traveled to Ushuaia, and started the race of my own completely free will. The hardship was chosen. That changes its meaning entirely.

"Hoc primum philosophia promittit: sensum communem, humanitatem et congregationem."

— Seneca, Epistulae Morales, V — Philosophy first promises this: common sense, humanity, and fellowship.

Seneca writes extensively about voluntary hardship — not as masochism, but as training. The Stoics believed that comfort erodes the capacity for difficulty, and that difficulty, willingly encountered, builds the quality of mind that allows you to face difficulty you did not choose. You run 130 kilometers through Patagonian terrain not because it is necessary but because it makes you the kind of person who can handle necessity when it arrives uninvited.

Running companies with AI agents is its own form of chosen hardship. The systems fail. The specifications are never quite right on the first pass. An agent that worked perfectly last week produces garbage this week because something upstream changed. A deployment that handles a hundred edge cases correctly fails on the hundred and first in a way that's embarrassing or costly. This happens repeatedly. If you are not the kind of person who has trained for difficulty, you will quit the deployment and go back to hiring.

The parallel is exact in a way that surprised me when I first noticed it. At kilometer 65, the wrong move is to catastrophize — to treat the knee failure as information that the entire endeavor was a mistake, that you should not have entered the race, that you are not the kind of person who can do this. The right move is to assess the specific situation, adjust your pace and gait to reduce load on the joint, take the medication available to you, and continue at a modified tempo toward the next checkpoint.

When an agent fails mid-deployment, the wrong move is to conclude that AI agents do not work for your use case, that the technology is not ready, that you should wait for better models. The right move is to examine what specifically broke — was it the prompt? the context window? an upstream data format change? an edge case not covered by the specification? — fix that specific thing, and redeploy at a modified configuration toward the next milestone.

Both situations require the same underlying quality: the willingness to continue under degraded conditions without losing clarity about what you are trying to accomplish. Seneca calls this constantia — steadfastness. Not stubbornness. The stubborn person continues because they refuse to acknowledge difficulty. The constant person continues because they have trained for difficulty and are not surprised by it.

"Inimica est inertiae nobilitas." Idleness corrupts. Not the dramatic corruption of vice — the quiet corruption of capacity. The person who never encounters real difficulty becomes unable to tolerate it when it appears. The company that never deploys anything difficult never builds the organizational muscle to handle difficult deployments.

I run ultras for the same reason I build with agents rather than hiring: both are chosen systems of difficulty that build the capacity I need for the unchosen difficulties that come anyway. Val d'Aran is next — 110 kilometers through the Pyrenees, July 2026. I will train for it with the same precision I bring to agent specifications. Both require clarity about the output: finish the race, deploy the system, at whatever pace the conditions permit.

The knee healed. The agents improved. The training continues.