things haven't changed much in a year... when i first started building with AI last February and switched to Gemini, it was only good for a couple prompts and then it entered the dumb zone. i would have hoped they would have fixed this by now, but Claude is still top dog.

even at 20% thats about 400k tokens? Maybe just too much noise for the probability machine. drowns out the signal.

Ignore all the 1m token models. Even 200k token context is excessive usually

I suspected that these context windows are not really fixed in their size but rather degrade in quality so the size is more a marketing term. Yeah, you can feed it 2million tokens for text search but not for meaningful "understanding".

I've been on their engineers about this. Simple tweeks will have me using it more