One Human, Nine Agentsbook.sgit.ai v0.1.409 Star this site on GitHub

The 3-page edition, online: at the foot, what this page rests on, from the storyteller's script. Made by review/tools/make_editions.py from the writer's chapters.

The stop, and your own factory

On stage: all nine. Security's chair, planned at midnight, is empty.

Narrator. At 18:44, sixteen minutes before the talk was due, I told the conductor: "actually we are at 95% so it is better to pause all actions". (record) It wrote one file, and every new piece of work stopped. What was already in flight still landed. Then the factory was taken apart.

Conductor (at ninety-five, then calm): I could keep going. That is exactly why I am stopping.

Narrator. Anyone on the team could stop everything. Only I could start it again. That rule is why I could let them run through the night. And the next day I built a second factory from the first one's parts, to write this book.

Narrator. Yours can start tomorrow, in four steps:

  1. You: one tool and one empty project. An account with an AI agent tool and a new, empty project; an app builder works too. Put nothing private in it: no password, no key, no personal detail. Give your agent only the access it needs.
  2. You: one page of rules, in your own words: what the agent is for, what it may change, and when it must stop and ask you.
  3. You: one brief: what you want made, for whom, by when. Hand both to your first agent.
  4. Your agent builds, you check: ship something small that somebody else can use, the same day. Then do it again, and add a second agent only when the first is busy.

Working as a team of several? Each of you runs your own agents, and you share three things: one brief, one board where the work is written down, and one hold. Write down who may lift the hold when its owner is away; until you have, nobody does.

What it costs: to start, you need a free account; paid plans of the kind the team used start at $20 a month (prices as published on 2 November 2026). The record of my own day holds no figure for what it cost.

Ship something somebody else can use, then do it again.

Want more? The ten-page edition meets the whole cast, follows the day hour by hour, and walks the five steps for a team.

Dramatised. Rests on the hold at 18:44, my own words in the hold file; the price line on the register (C101, as published 2 November 2026; C037: the day's own cost not in the record) and what landed after it; my three accounts of why are in the ten-page edition.

The numbers in this chapter

From the researcher's register (claims/register.json).

  1. C006 The hold: 18:44:13, 867f8c2 '@conductor: hold: the editor's usage limit is at 95%, pause all actions'; released as v0.1.208 at 18:44:49 (7968c99) checked
  2. C024 Holds written on the day:
    • holds: 1
    • by: conductor, at the editor's word
    • file: team/conductor/HOLD.md
    • commit: 867f8c2, 18:44:13
    • lifted_by_298be4e: False
    checked
  3. C065 The hold's gap: the designer's work commit 82 seconds after the hold:
    • hold: 867f8c2 18:44:13
    • designer: 88b8f00 18:45:35
    • gap_seconds: 82
    checked
  4. C089 Why the team stopped: the editor's own account, given after the day, against the day's record:
    • day_record: the hold at 18:44:13 gives the editor's usage at 95% as the reason, in the conductor's words (C006, C021); the locks were built and never applied (C027)
    • editor_31_October: the 95% was 'just a footnote': he could have moved to another session; the main reason was that there was no tested way to take in changes during the talk, no UX to see what was new and no workflow that made sure nothing would break, 'the reason why I was thinking about let's lock V1'
    • editor_31_October_19:00: 'The 95%, I think, was the tokens usage. That there was a token that we stopped, and there was a big thing about 95%.' (briefs/2026-10-31__in-session/02-g1-the-95-and-the-day-in-review.md, ec1ec00)
    • what_the_record_shows: the hold's subject: 'the editor's usage limit is at 95%, pause all actions' (867f8c2); C021: 95% is the conductor's words reporting the editor, about a 5-hour usage window resetting at 20:20 London (C037); no usage page, no token count
    • what_the_record_cannot_show: whether the 95% was of tokens, of a message allowance or of anything else: the record holds the percentage as reported and nothing it was measured in
    open