WEBVTT

00:00:00.059 --> 00:00:03.999
Now, let's start with prompt
and context engineering, and this in

00:00:04.019 --> 00:00:07.039
the end simply means
that you wanna write good prompts.

00:00:07.059 --> 00:00:10.980
Because with large language models,
whilst they can also deal with

00:00:11.079 --> 00:00:15.020
bad prompts and might still yield okay-ish
results,

00:00:15.079 --> 00:00:19.039
typically you get better results
if you have better input.

00:00:19.060 --> 00:00:22.459
Good input equals good results typically.

00:00:22.519 --> 00:00:26.019
And the input, a prompt,
is in the end really just a

00:00:26.059 --> 00:00:29.699
combination of having specific
instructions that

00:00:29.840 --> 00:00:33.520
clearly describe the task
or problem you want Claude

00:00:33.619 --> 00:00:37.599
code to work on or solve,
and providing relevant

00:00:37.659 --> 00:00:41.139
context and extra information without
providing

00:00:41.239 --> 00:00:44.479
unnecessary information, because
that can lead to worse

00:00:44.560 --> 00:00:45.779
output.