I ran into something interesting recently while using Claude AI to review some of my code.

During the day the responses were *fast*. I could paste a file, ask for suggestions, iterate quickly, and the workflow felt smooth.

But when I tried doing the same thing later in the evening — around *9 PM and after* — the experience changed a lot.

Responses suddenly took *much longer*. Sometimes it would sit there “thinking” for quite a while before returning the review.

At first I assumed it was something on my side:

- maybe my internet - maybe the browser tab - maybe I pasted too much code

But after trying a few times on different days, I noticed the same pattern.

### It’s Probably Peak Usage

My guess is pretty simple.

Evenings are when a lot of developers start working on:

- side projects - open source - debugging issues from the day - experimenting with AI tools

So tools like *Claude*, *ChatGPT*, and *Gemini* probably get a *huge spike in requests* around that time.

And when millions of prompts hit the system at once, response times naturally slow down.

### Why Developers Notice It More

The funny part is that developers probably notice this more than anyone else.

When you're using AI for *code review or debugging*, you're usually sending multiple prompts in a row:

- ask for review - ask for improvements - clarify something - test another idea

That workflow depends on *fast feedback*.

Even a small delay starts to feel frustrating when you're in that loop.

### Not Really a Complaint

To be fair, these systems are doing a massive amount of work behind the scenes. Running models at this scale is insanely expensive.

But it was interesting to notice how *time of day actually affects the experience*.

Now I’m curious if other developers have noticed the same thing when using AI tools for code reviews at night.

Is it just me, or does AI also have a *“rush hour”*?