1
00:00:00,005 --> 00:00:02,956
I'm going to show how I organize the work when I have three or

2
00:00:02,956 --> 00:00:07,082
more projects that are related, or that depend on each other.

3
00:00:10,030 --> 00:00:12,370
On the left are the three projects,

4
00:00:12,370 --> 00:00:15,770
the store API, the portal and payments.

5
00:00:15,770 --> 00:00:19,230
I add one project per repository, because the terminals,

6
00:00:19,230 --> 00:00:22,410
the stages and the rules are tied to the project.

7
00:00:22,410 --> 00:00:26,440
When I click payments, the Grid and the sidebar switch together.

8
00:00:26,546 --> 00:00:29,356
In the Flow I make one flow per goal.

9
00:00:29,356 --> 00:00:30,856
This one is for fixing the coupon,

10
00:00:30,856 --> 00:00:33,576
and in the selector there's the one for updating the docs.

11
00:00:33,576 --> 00:00:36,356
I keep them apart because each one has its own stages,

12
00:00:36,356 --> 00:00:38,874
and I'd rather look at one job at a time.

13
00:00:38,874 --> 00:00:40,414
To follow several agents I stay

14
00:00:40,414 --> 00:00:43,154
in the Grid, with the terminals side by side.

15
00:00:43,154 --> 00:00:46,074
When I need to read a long answer, I switch to Focus.

16
00:00:46,074 --> 00:00:47,634
The terminal gets big in the center,

17
00:00:47,634 --> 00:00:49,812
and the others stay in the strip below.

18
00:00:49,812 --> 00:00:52,892
When two agents are going to touch the same repository, I create a

19
00:00:52,892 --> 00:00:57,802
worktree, which is a second folder of the repository, on a separate branch.

20
00:00:57,802 --> 00:01:00,882
Each agent gets its own copy of the files.

21
00:01:00,882 --> 00:01:05,131
The card shows the line cupom-fix, worktree, one session.

22
00:01:05,131 --> 00:01:07,287
I choose the model by the task.

23
00:01:07,287 --> 00:01:09,927
For Codex I asked something short, to explain what

24
00:01:09,927 --> 00:01:13,877
the test script is for, and a smaller model handles it.

25
00:01:13,877 --> 00:01:16,137
A bigger model is slower and more expensive,

26
00:01:16,137 --> 00:01:19,302
so I keep it for tasks like finding the cause of a bug.

27
00:01:19,302 --> 00:01:23,471
On stages that matter I don't settle for the agent's answer.

28
00:01:23,471 --> 00:01:28,241
The card has the line Verification, npm test, detected.

29
00:01:28,241 --> 00:01:31,421
If a stage has this verification and the tests fail,

30
00:01:31,421 --> 00:01:35,060
it isn't marked ready, even if the agent says it's done.

31
00:01:35,060 --> 00:01:38,620
When a flow repeats, I save it as a routine.

32
00:01:38,620 --> 00:01:43,420
This one, Review and fix, has Claude reviewing and Codex fixing.

33
00:01:43,420 --> 00:01:45,660
And what applies to every agent in the project,

34
00:01:45,660 --> 00:01:49,930
like answering in Portuguese, I write once in the rules.

35
00:01:49,930 --> 00:01:53,580
Next, I show what a day with more than one project looks like.