Servers

Shared context diagrams

Five sketches of the problem in the product thesis. People relay what they know to each other, and each person's agent sees only what that person can reach. Each one loops while it’s on screen.

A team grows one person at a time

The March offsite starts with one person and grows to seven. Each card shows a person, the agents they chose and the accounts only they can reach. Without a shared place to work, every new person has to stay aligned with everyone already there. The story loops while it's on screen; pause it or step through the squares.

March offsite@march-offsite

Three days in Lisbon for 40 people in March. Budget capped at $40,000, and the venue must be step-free.

Separate silos become Unix users on one shared server

Each column is one person, their agents and the work only they can see. On a shared server, everyone gets a Unix user and a private home. Work that involves more than one person moves to /workspace, and ordinary Unix groups decide who can open each piece.

Each retelling loses detail

A client's requirements land in Sam's inbox and pass from Sam to Hannah to Mike. Each retelling drops a detail. With a shared server, the email stays private to Sam. Sam's agent writes the details the team needs to the shared /workspace folder on the server, and every agent reads those.

Why engineers feel this less

Mike already works this way. GitHub gives engineers one working record of code, issues, reviews and history, and every engineer's agent reads it. The rest of the March offsite has no equivalent, so context moves from person to person.

The relay grows faster than the team

Every pair of people has to stay aligned, so conversations grow with the square of team size. Connections to a shared server grow by one per person.