§ Recent writing
full archive →- May 2026Data
Why MCP Servers Are Going to Kill Traditional Data Pipelines (And That's a Good Thing)
MCP servers are replacing traditional data pipelines for real-time, context-aware applications. Learn why latency, rigidity, and maintenance costs make Airflow-style orchestration obsolete for AI agents while batch processing remains essential.
- May 2026Innovation
{
El Model Context Protocol (MCP) no es solo una moda: está redefiniendo cómo las startups diseñan sus sistemas de IA. Entendé qué cambia y por qué importa elegir bien el protocolo desde el día uno.
- Apr 2026Data
The Data Quality Metrics You're Using Are from 2020. Here's Why Your Pipeline is Broken
Exploring why traditional data quality metrics are insufficient for modern AI applications and suggesting practical steps to improve data quality.
- Apr 2026Data
Why Companies That Implemented Generative AI in 2024 Are Failing in 2026
Most generative AI implementations in enterprises didn't fail because of the technology — they failed because of something more fundamental: data.
- Apr 2026AI
The problem with AI isn't adopting it, it's governing it
Pilots die in the pilot phase. What separates a POC from a productive system is governance of the model, data, and consequences.
- Mar 2026Data
Three years teaching data at ITBA: what changed with LLMs
Teaching data in 2023 was SQL and models. In 2026 it's knowing when not to use an LLM. Notes from a program that gets rewritten every year.
- Feb 2026Innovation
Innovation isn't technology, it's organizational permission
An innovator's most powerful tool isn't what they prototype. It's the political cover they secure before testing it.
- Jan 2026Data
Democratizing data: the easy part and the real part
Dashboards for everyone is the easy part. The hard part is decisions actually being made with them.