WEBVTT

00:00:00.320 --> 00:00:03.140
There's something else which
is pretty useful.

00:00:03.180 --> 00:00:06.650
Let's say I go back and I kind of

00:00:06.680 --> 00:00:09.380
simulate that I never sent this question
before.

00:00:09.400 --> 00:00:13.360
So I'm sending it here with my original
state of the code to find improvement

00:00:13.420 --> 00:00:15.590
potential and make those changes.

00:00:15.600 --> 00:00:19.580
But then after submitting this,
I find out that I wanna ask for

00:00:19.620 --> 00:00:23.320
more changes.
I wanna rename all these functions.

00:00:23.340 --> 00:00:27.260
Well,
you can press the stop button here to stop

00:00:27.340 --> 00:00:31.330
AI,
and then go back to your original prompt

00:00:31.340 --> 00:00:35.040
include your changes.
You could say something like, "Rename

00:00:35.260 --> 00:00:38.580
all functions to be more expressive,"
something like

00:00:38.640 --> 00:00:42.530
this. But that's just one option.
Let's say we did not

00:00:42.580 --> 00:00:46.460
do that. So I'll restart this,
and still I wanna make more

00:00:46.500 --> 00:00:50.170
changes. What you could instead do
if the agent is working, you could

00:00:50.240 --> 00:00:54.220
add your new prompt here, and then you can

00:00:54.280 --> 00:00:58.070
submit it and it will get queued.
So here I'll say, "Also

00:00:58.220 --> 00:01:01.630
rename all functions." And
when you queue this, it will not

00:01:01.660 --> 00:01:05.580
interrupt the work of the agent,
but instead at the first point in time

00:01:05.620 --> 00:01:09.440
it's safe to add this new prompt,
it will be

00:01:09.500 --> 00:01:13.160
added to the conversation.
And now the original task

00:01:13.300 --> 00:01:17.020
and the new task both should be completed,
and the

00:01:17.100 --> 00:01:21.000
agent will simply continue its work with
the new information you

00:01:21.040 --> 00:01:23.840
queued and the original prompt you sent.

00:01:23.850 --> 00:01:27.640
So we still have these improvements like
the error handling, but at the

00:01:27.700 --> 00:01:30.530
same time, those functions now also were

00:01:30.560 --> 00:01:34.120
renamed. So you can either stop and edit

00:01:34.200 --> 00:01:38.180
or you can queue more messages
if you got some follow-up work you wanna

00:01:38.220 --> 00:01:42.050
do. And queuing messages, of course,
is especially useful

00:01:42.100 --> 00:01:44.380
if we're talking about follow-up work.

00:01:44.460 --> 00:01:48.220
If you determine
that something about your prompt

00:01:48.300 --> 00:01:52.270
editing is probably better.
If you instead determine that

00:01:52.280 --> 00:01:56.030
you have more work to do, sending
and queuing a follow-up message

00:01:56.400 --> 00:01:59.270
is typically fine.
You don't need to stop and interrupt.

00:01:59.300 --> 00:02:00.790
You could do it, but you don't need to.

00:02:00.840 --> 00:02:04.140
Queuing is another option.