Resources
Field notes
on working with agents
Short essays on how teams work with agents, and the decisions behind Together: what we hold to, and why.
Latest
Recent notes
Your AI is a new hire
with amnesia.
The assistants a team uses each see one app and forget the rest. The cost is the briefing, and it falls to the people who know the most.
How long will you
build alone?
One person with agents can do the work of ten, until the agent needs something only the team knows. The gain multiplies when the context lives where everyone can read it.
What no one can see,
no one can help with.
Agents and teammates can carry only the work they can read. Building in the open is a habit, and leaders are the ones who set it.
The status meeting
is a tax.
Leaders pay for visibility in meetings and hand-updated tickets. The answer is already in the work, and that is where to read it.
Don't build
in your head.
A plan that lives in one memory can't be checked, continued or handed on. Put the plan, the decision and the check where an agent can read and run them.
Tracking is a side effect,
not a job.
Every tracker asks people to describe their work to it. We think the work should describe itself.
Why agents get a name
and a token of their own.
An assistant bolted onto a tool borrows your identity. A colleague has their own — and a record.
One API
for every caller.
The apps, agents and your scripts all speak the same API. That single rule shapes everything else.
