Tfw you think one of your ai agents is the cause of you getting locked out of all your company accounts…but in reality a new IT security policy revokes all your oauth for the automation emails they setup up and inadvertently breaks 14 ai workflows in production. #thestruggle
I was getting really pissed @claudeai after opus 5 & spending day in & day out dealing w/ a novel of meta analysis of each session interaction + usage limits + lost context + slop code. Opus 5.5 won me back…insane how much more token efficient & better it is @boris_cherny
Hope you'll tune in to my conversation with @thevelocityroom next month! We'll dive into the challenges of developing an AI strategy & building agentic workflows in the world of healthcare
https://t.co/valyPU4LFu
@Alex1here Nice. Gotta love these little demon agents. They’re like 5 year old Einstein’s that happen to be drunk and blink and running around with a loaded gun 😂
@Im_IrushiK They’ve extended the usage boost like 10 times already. Doubtful they’ll bag it given the price difference between sol 5.6 and Fable/Opus 5
Originally started working on this game when codex was first released. It was really janky. Image gen wasn’t even at the level of the original flappy bird. This is a great example of how far ai has come in the past 6-9 months. https://t.co/6diagpocQA #codex#aisoftwaredev
@bcherny@claudeai ya’ll really need to figure something out re: usage limits. This is getting out of control. Yes, you can run agentic loops & graphs. No, this isn’t viable for every business challenge or project. #ailimits
Does anyone else find opus 5 bordering on unusable? Yes, it’s very good for complex tasks and managing sub agents. Everything else? It’s basically 🗑️ #provemewrong#claude#opus5