Thats a lot of tokens.. .5.8 BILLION! Yea if I used that many I'd probably want a monitor of some kind.. wow.

Yeah, I make sure I get my $200/m worth

So Anthropic got rid of the peak sessions using more 5hr session tokens, and increased 5hr session usage when they contracted with spacex to use their datacenter.

It's nice, I can go most of the day doing a crap ton now in multiple sessions and not hit my 5hr session limit

I noticed this, switching back to claude as my main gpt

https://www.anthropic.com/news/claude-opus-4-8

Is it me or versions coming out faster and faster.

Seems so

I haven't used 4.8 yet.

Whoa... ya this is neat.
So I am doing a clean up on an API I have that has a ton of deprecated code. But still used heavily in another application.
I am doing a deprecation cleanup pass, the updating the usages in the other app. 4.8 spun up a bunch of agents for exploring automatically.
in 4.7 you'd have to tell it manually to do that or use superpowers and it would too. but this seems better.

I saw they had a new feature.. ultramode? or something like that.

ultracode is the new effort, no idea how it does

From Reddit:
```
Ultracode uses agents one tier lower than Max but in a collaborative way that may outweigh Max in trend result against just using Max.
```

no opus 4.8 yet, but QWEN is moving up!

How do you guys Keep your input tokens on check with medium code bases?