I have uploaded a recording of a separate Montezuma's Revenge run to YouTube.
This run took a whole lot longer than the other one. The timestamp of it discovering the final room is 1:40:30 youtube.com/watch?v=AeWLyBcP…
GPT-6 Astra (low) has beaten the Atari game Montezuma's Revenge, in real time, with a basic harness
This presumably resolves the Metaculus question "When Will Weakly General AI Arrive?"
DH3 on the Artificial Analysis Image Arena appears to be Seedream 4.0 from ByteDance
SoTA image generation and editing
fishy business retweeted
gpt-5-reasoning-alpha-2025-07-13
h/t @swishfever
if you work at a big lab and want to shitpost or post leaks check this out gossiping.ai/
fishy business retweeted
"Claude 4 is here" - "Try Claude Sonnet 4 and Claude Opus 4 today"
"Try Claude Sonnet 4 or Claude Opus 4 for Anthropic’s smartest models yet."
"Not intended for production use. Subject to strict rate limits"
"show_raw_thinking" / "show_raw_thinking_mechanism"
(not available yet - discovered in the feature gates configuration)











