WEBVTT

00:00:00.500 --> 00:00:02.282 align:center
Okay, here's a quick demonstration of

00:00:02.343 --> 00:00:03.224 align:center
Kawa code.

00:00:03.244 --> 00:00:05.006 align:center
I started with MLflow, this issue,

00:00:05.046 --> 00:00:06.368 align:center
they posted on the website, loaded

00:00:06.428 --> 00:00:08.491 align:center
into VS Code, and then I

00:00:08.551 --> 00:00:12.136 align:center
started with a cloud code just

00:00:12.176 --> 00:00:13.678 align:center
by itself to help me fix

00:00:13.718 --> 00:00:14.259 align:center
this issue.

00:00:15.120 --> 00:00:16.842 align:center
And it fixed it, but in

00:00:16.902 --> 00:00:18.084 align:center
a very marginal way.

00:00:18.104 --> 00:00:19.286 align:center
There were plenty of other scenarios

00:00:19.326 --> 00:00:20.808 align:center
that didn't fix, and it consumed

00:00:20.828 --> 00:00:22.830 align:center
71% of the session tokens.

00:00:23.331 --> 00:00:25.534 align:center
Then I used it with Kawa

00:00:25.574 --> 00:00:27.216 align:center
code, and I started the same

00:00:27.457 --> 00:00:29.840 align:center
prompts, everything, and...

00:00:29.870 --> 00:00:31.692 align:center
it went through the whole thing

00:00:31.732 --> 00:00:33.474 align:center
it built some tests and it

00:00:33.935 --> 00:00:35.576 align:center
did generate the correct solution this

00:00:35.616 --> 00:00:44.006 align:center
time for only let's see there

00:00:44.046 --> 00:00:47.089 align:center
you go 59 percent and here's

00:00:47.650 --> 00:00:49.912 align:center
an AI analysis a comparison between

00:00:49.952 --> 00:00:51.394 align:center
the solution we found in core

00:00:51.414 --> 00:00:53.276 align:center
code and the solution that the

00:00:53.316 --> 00:00:56.039 align:center
maintainer later accepted from one of

00:00:56.059 --> 00:00:56.700 align:center
the contributors