After the GPT-5-Dud, I think I understand why Satya Nadella said, "All I know is, I'm good for my $80 billion."

Congratulations on this great release Nous Team. Teknium (e/λ) Roger Jin Jai dmayhem93 emozilla Joe Li NobodyExistsOnTheInternet Shannon Sands mephisto I think this is the first time I'm seeing authors' Twitter handles in place of email IDs. These "tpot(s)" are becoming

$1 Trillion value added in the past 5 months! And also - the 14% stake in Anthropic and some stake in SSI Inc. - Major non-search independent verticals like Maps, Gmail, Drive/Storage & Photos/Media hosting. - Open Platforms like Chrome & Android and solid consumer hardware

This is probably one of the most creative and useful user interface/surface I've seen in the context of image gen. IMO, Canvas is the most useful surface/interface architecture after the "vertical chat" surface/interface!

//thousands of concurrent agents each working for up to 12 hours at a time// Congratulations Christian Szegedy, Jesse Michael Han, and the teams of Math, Inc. & Morph! The bets made by the terrific-trio frontier labs on "multi-agent" and "extremely long & parallelizable TTC

Since this is a custom model/system, the infra credit should go to Fireworks AI Great job Lin Qiao, Dmytro Dzhulgakov, Benny (Yufei) Chen, and team! cursor.com/security

I believe we've arrived! With 3.5 months left in 2025, it appears we're only about 1-2 months behind the prediction targeted in ai-2027.com It's not as sci-fi as anticipated, but great progress nonetheless.