OPEN NOTEBOOK · 15 entries
Gnoses
Each gnosis carries four visible fields: maturity, confidence, sources and backlinks. If I say something here, you can see where it came from and how much I would bet on it.
AI proficiency is now competitive edge
A 3-minute crisis solved with AI exposes who leads without understanding the work — and why that just got expensive.
Ethos: the layer that precedes retrieval
In Gnostikon, ethos is mandatory: principles, rules, limits and style are injected before any retrieval. No ethos, no action.
Strange loops and the substrate of identity
Notes on Hofstadter and the hypothesis that the self is a recursive pattern of self-reference — not a substance.
Clostera & the AI optimization era
Technical analysis of BaseModel.ai's open-source clustering library. Why elegant math + optimized hardware still wins.
Alignment as an epistemic problem
The ethical formulation of alignment is parasitic on a prior epistemic formulation: how to prevent the system from lying about what it knows.
The era of cheap AI is over
Compute is the bottleneck. Bottlenecks become price. Whoever used the cheap window to integrate and redesign now operates at a different level.
Chomsky vs Pinker — the grammar axis
Quick map of the internal dispute between universalist generative grammar and language as evolutionary adaptation.
Aletheia — the validation layer
The Aletheia layer carries coherence, confidence, conflict, gap and paradox. Its function is to make visible what the system does not know.
VTEX Day 2026 — Search & Personalization as a strategic asset
Panel on how sponsored media, algorithmic relevance and personalization redefine the role of search in retail.
Hacker mindset — mental laziness as adversary
Five operating principles derived from my trajectory. Non-conformism, autodidactism, truth-seeking, happiness through problems, ownership.
Did the junior dev die? No — the definition changed.
The junior role as conceived in the last 20 years is fading. What remains is a career that starts thinking about product and system, not small code.
Thousand Brains — cortical reference
Hawkins proposes the neocortex is thousands of parallel reference models voting. How this reshapes what we mean by "understanding".
Gödel — what formal systems cannot
Re-reading the incompleteness theorems through what LLMs inherit: the impossibility of total self-validation inside the system itself.
379 hours — source code of human intelligence
Audible 2025 wrap: why I studied Chomsky, Pinker, Hofstadter, Strogatz, Nietzsche and Aristotle instead of prompt engineering.
Nike Chile case — applied behavioral AI
+26% orders, +30% demand, +30% sales/ticket via behavioral AI communication and real-time product substitution.