Atelier-Evaluation

All right, let's kick off some work.

I have my project in the bottom left here.

I'm going to take this game engine project and I'm going to tell it to take the fighting game and turn its background into animating rainbows.

That's going to create its own workspace here at the top.

And I'll pause the recording And we're back.

The agent finished over here.

And you can see that we're playing this game and, the background is all rainbows.

Now the cool thing is not that it's all rainbows, but it's that the agent did all the work to present the game for me.

I didn't have to go and figure out sort of like how to run it, how to open a browser, which, you know, which workspace it was.

It just, it just opens the remote development browser, for me with the right content so that I can immediately form an opinion on the task it performed and I don't have to spend any mental cycles on anything else.

Let's do the same thing with another project here, the math tutor.

I'm going to send this prompt over there and I'll pause the recording.

And we're back.

The agent finished.

I'm going to go here.

And you can see also in this project, which is a math tutorial, it changed what used to be a red parabola to a rainbow one here.

And again, that is not the special part.

The special part is that the agent decided to use the remote development experience and to open the browser in it and to not open the browser at the beginning of this lesson.

Like it opened the browser at like chapter eight because it knows that that is the chapter in which this red parabola, appears.

So when I arrive at this workspace, I don't have to do any, any steps.

I can just immediately observe the work of the agent.

Okay, let's do one more.

I'll go here to the PI coding agent and say, change the settings screen in PI to be all animated rainbows.

And I'll pause the recording again.

And we're back.

Let's take a look at what it made.

look at that.

We have the PI settings screen all animating rainbows.

Now PI.

The project you see here is a coding agent.

It's a terminal coding agent.

So this is not being shown in the browser, but it's shown in terminal, window.

And here the agent decided that for me to evaluate animated settings page, the best thing would be to start PI in a TMUX session to navigate that TMUX session to the settings screen.

Because this is not normally, like this is not normally where you start off in PI.

You start off here, go there like this, and to prearrange it at the settings screen and then use Atelier's development environment to attach a terminal window to that same TMUX session.

So, same as with the previous two examples, I have to do zero work to be able to evaluate the thing I see here and it allows me to make my decision.

If I actually like what I see, I would follow up by actually going to the review pane and take a look at the code.

Depends on what I'm doing, how deep I would get into that or not.

And then from here I can also do things like select this and put a comment here.

No, not like that.

Maybe do a few more comments and when I'm done with that, I'm just going to hit this button and they're going to end up in the Agent composer and I can just send them for processing.