GPT 4.5 initial impressions:
- explicitly not a reasoning model, clearly OAI pushing the boundaries of pre-training/unsupervised learning (fattest model by far)
- we're clearly going to run into a limit (data/compute wall), with test-time scaling being the new battleground
- gpt4.5 = street smart, o1/o3/3.7 = book smart
- jfc what are they smoking with this pricing

cofounder Ohara.ai — mini apps! building software generation on @base

just tried the new coffee place on 5th. barista named sam made the best latte i've had in ages. def going back! anyone else been there yet?