What We Actually Work With
A lot gets said about models. Hardly anything about the surroundings. Yet in daily work the surroundings are what decide. What is described here is not a product and not something we sell. It is the answer to the question we get asked most in conversation, mostly by people who work with AI themselves: how is this set up at your place. The Editor Is The Stage The work happens in a normal code…
The most important aspect of how work is conducted in this AI-first digital studio is the surroundings, rather than the models themselves. Assistants operate within a standard code editor, granting them access to files and the ability to execute commands and interpret results. This contrasts significantly with a chat window, which only sees what is pasted into it.
Two separate model families are employed side-by-side, rather than just one, as different models can detect varied errors and blind spots in the work. Recipes are used for recurring procedures, recording steps such as deployments or server checks to prevent knowledge loss. Strict "hard stops" are implemented to prevent certain actions, providing a technical barrier instead of a verbal admonishment.
These stops are crucial to prevent errors that could have widespread impact, such as overwriting a server or deleting data. A key component underpinning this entire system is the memory, which retains decisions, mistakes, and patterns accumulated over time across sessions. This memory allows the system to warn about past mistakes, rather than each person having to individually remember them.
The overarching principle is that the assistant should be an integral part of the work environment, review should be conducted by a different individual than the one who created the content, and any previously identified issues should have a firm constraint in place to prevent recurrence.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
