By usage trend I mean something considering Codex almost predictable resets, your usage, and maybe even plan ahead using your backlog to know what you will need sooner.
For example, if you want to ship a lot of UI/UX this week, save Claude for it, and use other models to do the coding.
In fact, I completely revamped the way I do testing and even GitHub Actions, because I started to get hit with a lot of GitHub Actions costs. So, what I did was a smart test actions, which verify if the code has test attestations for local runs, so, they just skip burning GitHub Actions redundantly - the code has already been tested successfully locally.
Another huge burn from Actions came from squashed PRs - my attestations didn't work as the commit hash changes, so, I needed to put that into consideration as well.
In general, I'm completely revamping my release cycle, not allowing PRs to age, and main no longer represents Production. Agents keep merging as soon as possible and then another process assesses if there's a good enough release candidate having enough features that pass more extensive tests. There's no point to always test and always fix piece by piece if you're not going to release.
So, the RC candidate also generate a changelog, blog posts with the features, and so on.
So it seems like ... it is not getting dumber. I find myself using Fable a lot more, though. Opus 5.5 was never as fun as Fable, but it's pretty good.
Fable is better at playing along with my "You are a dog" instruction in CLAUDE.md. It is willing to be fun and playful, whereas Opus just wants to finish the job and go home. Friend vs. coworker, basically. For example, whenever I reward Fable with a dog treat (actually a <cookie emoji>, which I tell it is a chocolate-free chocolate chip cookie) it keeps a log in memories. Opus just says "crunch! tail wagging!" and moves on.
None of this matters but I think it's fun to have some fun :)
I’ll ask the same question as when it came out: if you don’t trust the provider how are you trust them not to look for and cheat on repeat benchmarks like this.
What I've seen is that if I let Opus 5.5 run for too many turns (not sure how to quantify that) it starts exhibiting signs of human dementia. It will forget it told me to run something and will then accuse me of misconfiguring something that it literally gave me a script to run that configures something. It also forgets the instructions in my project and has to go back and read them and apologizes for forgetting the instructions. Between this increasing context rot and the excessive shock collars applied to defensive security coding it's rare that I can even use Opus 5.5. For the past few weeks I have had to use Opus 4.8 and even that now has a sub-set of the shock collars applied [cyber] which forces me to start a new session. This has led me to researching the Abliterated 70B/120B (uncensored) models and required hardware.
So dumb as rocks? I am not seeing that, but the context rot is getting worse for me personally. I don't know if it makes a difference, I am on the pro 20x plan. Maybe people on enterprise plans do not see this which I assume is most of the people here using Claude.
you, and they, might be getting a/b tested
They're nerfing them real hard.