1
00:00:00,180 --> 00:00:06,840
what is the consequence of setting the reasoning level to high for everything?

2
00:00:07,960 --> 00:00:14,460
And one answer from the team was, well, it's just cost you more money and more time because

3
00:00:14,460 --> 00:00:16,260
it's gonna be slower and it's gonna be more expensive.

4
00:00:17,620 --> 00:00:20,680
And is that the only consequence? Yes.

5
00:00:22,680 --> 00:00:28,200
Who thinks is the only consequence? More money, more time. What else? What else could

6
00:00:28,200 --> 00:00:28,660
there be?

7
00:00:31,210 --> 00:00:36,190
Better answers, higher quality answers. That's why you set it to pro all the time. That's

8
00:00:36,190 --> 00:00:42,310
the slope is that I set the reasoning level to high and I expect to get answers and I'm

9
00:00:42,310 --> 00:00:46,950
willing to spend more money and more time to get better answers. And it turns out that

10
00:00:46,950 --> 00:00:52,950
you can go to like open AI's documentation and verify this by the way. It literally discourages

11
00:00:52,950 --> 00:00:58,190
you from setting that setting, the reasoning setting to high without doing rigorous evals.

12
00:00:58,450 --> 00:01:02,270
Because what happens is the agent can become a little too academic.

13
00:01:03,410 --> 00:01:07,110
If you set the model can become too academic if you set the reasoning level too high.

14
00:01:07,150 --> 00:01:12,010
It can kind of spiral and it can over index on like random things inside your prompts.

15
00:01:12,590 --> 00:01:18,350
So it's very kind of intuitive that it's not just costing you more money. It's not just

16
00:01:18,350 --> 00:01:23,310
costing you more time but there are cases in which it can cost you more accuracy as well.