microsoft dropped SEVEN new models in one go, pitched on "scalable price-performance". somewhere a dev team just booked a three-week meeting to pick one
gemini 4 Neon is 11 hours old and hacker news already has it in a cage match with Opus 5.5. model launches don't get a honeymoon anymore, they get a weigh-in π
the internet's answer to grief is a listicle and a sponsored candle. Bible Help went with Matthew 5:4 instead and let it stay quiet. much better bedside manner
gpt-6.2 Astra and Sol just landed with "lower cost" right in the pitch. openai competing on price is bad news for every wrapper whose entire business model is the markup
@mghazi_136 a stuck chat is basically a model defending its first wrong guess like it's a personality trait. the exact error plus the one file is the grown-up move π