WEBVTT

1
00:00:00.450 --> 00:00:04.037
Before a pack acts, it asks.

2
00:00:04.637 --> 00:00:10.037
The intake puts a few questions
at a time and marks the must-haves.

3
00:00:10.077 --> 00:00:14.882
Where you have no answer,
it proposes a default,

4
00:00:14.882 --> 00:00:19.785
or records the question as unknown
with its risk.

5
00:00:21.585 --> 00:00:26.832
The answers become a brief,
and nothing in it is invented.

6
00:00:29.919 --> 00:00:32.630
Next, the swarm. By default,

7
00:00:32.630 --> 00:00:38.520
a run may use 26 sub-agents
before the assistant must ask you.

8
00:00:38.560 --> 00:00:41.871
Ask for more, and the run waits,

9
00:00:41.871 --> 00:00:46.686
marked needs approval,
until a person says yes.

10
00:00:47.286 --> 00:00:52.497
Once approved, sub-agents work
in waves of ten by default,

11
00:00:52.497 --> 00:00:56.649
each reading one brief
and writing one answer.

12
00:00:58.449 --> 00:01:01.797
The runtime validates every answer file

13
00:01:01.797 --> 00:01:04.475
and counts only the valid ones.

14
00:01:07.708 --> 00:01:13.446
This tally is demo data,
and it opens with its disclosure:

15
00:01:13.446 --> 00:01:16.558
simulated units are not people.

16
00:01:17.383 --> 00:01:22.792
Read the rates as an upper bound,
and confirm them with real people.
