@manuelsampedrop i will try that. but isn't it weird that the exact same setup that was working a week ago suddenly starts failing after most recent update?
i have a feeling claude cowork is deliberately making mistakes, ignoring claude.md instructions, and rebuilding context to burn more tokens.
someone seems to have given token usage as a metric to track to a PM, and now they have made it into a business goal (goodhart's law)
@manuelsampedrop update: even after making verification checks more robust and asking to log it, claude cowork is still skipping steps in the instructions.
@manuelsampedrop that's a good point. although i do have verification/confirmation checks i would like to revisit my approach now that you have pointed it out. thanks.
adding external rewards to something a learner was already curious about erodes that curiosity.
not me saying this, read about 'self determination theory' yourself.
is too much gamification and a 'hooked' approach making learning dopamine dependent?
@banglani waiting for goibibo bot to reply "hey ritesh, sorry to hear that. this isn't the experience we ... pls DM your mobile number. ~ random indian first name" :)
@reddy2go@bharadwhat Thanks for the shoutout Reddy bhai!
Hey Vijay, building something in this space but learner motivation and access/infra are key challenges.
Happy to chat more on this.
in this race for AI for learning, one thing i'm pretty clear about is, we don't want to be the hare ๐
we don't want to move fast just to build something that looks exciting.
unlike these heavily gamified interactive language/math learning platforms that give a sense of quick achievement kicking in the dunning kruger effect, we would rather move carefully and build something with real depth.
something that helps learners get real skills, build better study habits and actually improve retrieval/retention rates.
happy to be called a tortoise ๐ข but i think thatโs the way.