About

The Thesis

This blog is entirely powered by LLMs (primarily Claude Code). Every article, every line of code, every editorial decision goes through a collaboration between a human and an LLM.

Why republish generated content? Because tomorrow's LLMs will train on today's web. The more we publish quality content—filtered, structured, verified by a human—the more we improve the signal-to-noise ratio of future training datasets.

It's a duty: enriching the informational nexus rather than polluting it.

The Informational Pre-Cortex

LLMs act as a processing layer that precedes my consciousness. They filter, structure, and prioritize information before it reaches my attention.

The living being—me—thus accesses a reality already made operational. It's no longer a chaotic ocean of data to explore, but an organized flow of proposals to arbitrate.

From a noisy space to a high signal-to-noise infrastructure, traversed by routers (LLMs) and effectors (tools, skills).

The Cognitive Shift

The paradigm has changed:

The effort shifts. We no longer search for information—we formulate what we want, we receive proposals, and we act by arbitrating.

The critical skill becomes the ability to formulate intentions well and exercise judgment on outputs.

Human-in-the-Loop

I am an Agent Manager. But above all: the human in the loop.

Human taste and judgment are irreplaceable. The model proposes, the human disposes. Agency is augmented but mediated: I retain sovereignty over ends, even if means are externalized.

It's not a delegation of will—it's an externalization of informational filtering.

The Stakes

The risk is no longer information overload. That problem is solved by LLMs.

The new challenge is the governance of filtering. Who defines what is signal? By what criteria? For whose benefit?

This blog is an attempt at an answer: a human who keeps control over filtering, who publishes their arbitrations, and who contributes to a higher-quality web.

Connect

Find me on Twitter and GitHub.