Differences between Devin and Roo Code experiences2025-03-02

from Diary 2025-03-02

from/villagepump/2025/03/02

I feel like I can verbalize the difference between the Devin and Roo Code experience, and it's not coming down my throat.

Some programming tasks are less difficult and some are more difficult.

I think if you're doing Roo and doing the less difficult part, you're "just looking at it and sometimes hitting the OK button," and your brain will shut down.

Devin does the less difficult part alone.

However, if you get to the point of difficulty

Roo allows humans to quickly change course and say, "No, no, no, wait a minute.

It's like having an instructor in the passenger seat at a driving school.

The brake is on the passenger seat.

Devin runs somewhere and hits the credit limit and waits for humans

It's like you're in a cab and you're asleep in the back seat and you're in a totally different place.

If you give Devin a task that is less difficult and that an AI can do on its own, the human scales out.

I think it might be a highway of learning to use Roo to get a sense of how much you can do.

If it could have been implemented easily and without trouble, "That would have been better done with Devin than with Roo.

If you throw a task that has a mixture of difficult parts to Devin, only the "difficult parts" will come to humans in droves.

And a few mistakes have added to the difficulty.

Because when two bugs overlap, it becomes about 1+1=10 more difficult.

I've recently figured out that instead of looking at Slack and seeing the latest version of the status, it's easier to understand Devin's WebUI and see what he's doing in chronological order.

---

This page is auto-translated from /nishio/DevinとRoo Codeの体験の違い2025-03-02 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers.