Elevated errors on Claude Opus 4.6

21 pointsposted 12 hours ago
by rob

8 Comments

ed_mercer

11 hours ago

How is this still a problem? This has happened so many times now.

mpalmer

11 hours ago

I mean... it's not a microblogging service, or even Github. Think of the resources you are marshaling and tapping into when you request Opus inference.

Think of the statefulness of the systems necessary to manage prompt caching.

Not to mention all the enterprise SLAs they have to meet before they serve traffic to gen pop.

Additionally:

- I believe Anthropic is running a planned promotion where short-term limits are doubled? Could be off on the dates though.

- It's two weeks before EOQ, and everyone's no doubt modeling and planning EVERYTHING last minute, because with Claude they can

- It is valuable infrastucture and thus a potential target of attackers

csto12

10 hours ago

Sure, but they get less sympathy when a lot of their high profile employees talk about using Claude to write 100% of their code and yet Claude Code has loads of issues and their services go down every 10 minutes.

mpalmer

10 hours ago

Not my experience so your exaggeration doesn't really help make your point.

poemxo

11 hours ago

9 retries on easy prompts, all morning and afternoon

bluesnowmonkey

11 hours ago

This one is taking them a while to fix.

tills13

11 hours ago

Have they tried asking Claude to fix it no mistakes?