§ Blog and press
What we learn building collaborators
Working notes and announcements. What a digital collaborator is, how a role is built for one concrete process, and why everything it does ends up written down.
25 posts
Article
When an AI project dies, it is almost never because of the AI
When an AI project falls apart, everyone blames the model. Sort the causes by origin and the pile AI brought is the smallest. The rest is the same old management.
Article
Nobody installs AI and walks away
Software stays still until someone changes it. AI starts drifting the day you stop watching it, and it fails silently. That is why you don't install it and walk away: you operate it.
Article
The license is the cheap part
The license price is the smallest number you will pay for AI. The big money is in the team that runs it, the maintenance that never ends and the token consumption.
Article
Measure AI's return against a role
AI's return was impossible to isolate while it did scattered tasks. Once it takes on a complete role, you compare it with something you already pay for: what that role costs you.
Article
Your data stays home
When legal says the data can't leave, the AI project gets shelved. But where AI runs and what data it touches is designed. That restriction is the specification, not the verdict.
Article
A task is not a role
An AI agent and a digital collaborator are not the same thing at different prices. They work at different levels, and mixing up the name is the fastest way to set the wrong expectations.
Article
Harness: What surrounds the model
The model's intelligence is not what makes an AI work. What makes it work is the harness: six decisions a company makes for a digital collaborator, the same ones it makes for a person.
Press
This is Yunt, the AI that aims to solve and simplify work in companies
El Dínamo covers how Yunt was born and the five requirements of a digital collaborator: a role, a boss, evidence, audits and accountability.
Article
Astra, and why we are AI agnostic
OpenAI shipped GPT-6 Astra; the next is weeks away. AI agnostic means swapping the model is one line, and your handbook, data and records stay yours.
Page 2 of 3








