Really cool demonstration of gambler’s fallacy. DO NOT READ STEP TWO UNTIL STEP ONE IS COMPLETE.
1. Play Aaronson Oracle
2. Play Aaronson Oracle but this time consciously repeat inputs (ie input multiple left arrows in a row)
It sounds ridiculous but there is a pretty significant anti-AI sentiment developing amongst the masses
If the tech community doesn’t start addressing this issue right now, this will escalate quickly
We need to start working on some sort of UBI / UBR
ChatGPT advanced voice mode is insane.
Ask it for:
Singing
Humming
ASMR
Whispering
Different tones (yelling, quiet)
Different pitches (high, low)
Nailed every single one.
Quick GPT-5 review: most users will notice a big jump due to the model routing, resulting in their first experiences with reasoning models.
Small/medium step up from o3 based on use cases.
Still an impressive, frontier model, but not groundbreaking.
gpt 5 is the most bipolar model I’ve used. Some responses are frustratingly dumb, some responses save you a few hours.
Based on my experience, the more constrained a tasks requirements are the better it does. It is horribly uncreative.
I’m resisting updating my phone to liquid glass, however it looks absolutely gorgeous on an iPad, while still being fully functional. Perhaps I should give it a shot, but I worry my phone will not be able to handle it.