Notes from a distributed team that does not have an office

We get asked how the team works more often than we expected, usually by people evaluating whether to join. The honest version is more mundane than the usual remote-work essay and probably more useful.

Writing is the default, not a fallback

Almost nothing important at Yaju is decided in a conversation that leaves no trace. Not because meetings are bad, but because a distributed team spread across time zones cannot rely on everyone having been in the room.

The practical form of this is that proposals are written before they are discussed. A short document, circulated, commented on asynchronously, then a call only if the comments reveal genuine disagreement. Most of the time they do not, and the call does not happen.

This is slower for the person writing and much faster for everyone else, which is the correct trade in a team where the reader-to-writer ratio is high.

Decisions need a named owner

The failure mode of asynchronous work is not conflict, it is drift. A question gets discussed, everyone contributes, nobody decides, and it resurfaces three weeks later with the same arguments.

We handle this the same way we handle agents: everything has a named owner. Not a committee, a person, whose job is to decide after listening rather than to build consensus. Decisions are written down with the reasoning, so the next person to ask can read why rather than reopen it.

Overlap hours, not core hours

Rather than a fixed working day everyone must be present for, we hold a smaller window of overlap and protect it. Outside it, people work when they work.

This only functions because of the writing discipline. Without it, the overlap window becomes the only time anything happens and the team quietly reinvents an office day with worse ergonomics.

What did not work

Trying to reproduce informal office contact through scheduled social calls. It felt obligatory, attendance drifted, and the people who most needed connection were the least likely to attend.

What replaced it was less structured: pairing on real work more often than strictly necessary. The contact happens as a side effect of doing something together, which is how it worked in an office too.

We also over-documented for a while. Every decision written at length, including ones nobody would ever revisit. That produced an archive people stopped reading, which is worse than a smaller one they trust.

Why in person still matters to us

We prefer working together in person when it is practical, and we say so rather than pretending distribution is universally superior. Certain work, particularly early design and difficult disagreements, goes better face to face.

The position we have landed on is that remote is the default and in person is a deliberate choice for specific work, rather than the other way round.

If this sounds like your kind of thing

The Careers pages cover open roles and how we work, and the Scholars Program is a remote, part-time way in for people who want to work on agents with a mentor. The Open Development Community is free to join and is where a good deal of the work happens in the open.