FRAMEWIREIndonesiaUpdated Oct 1Live wire
0:00 / 0:00

GPT-6.1 Sol Ultra vs Opus 5-5 Ultracode.

Two AI agents independently controlled separate computers through computer use and joined the same multiplayer match. Both perspectives in one video. Open-source project coming soon.

Mad Cat LabsOct 13
0:00 / 0:00

Yo AnthropicAI why are you routing opus 3 to the sonnet 5

Anon🇨🇭Oct 11
0:00 / 0:00

60% cheaper cache reads.

Higher elo than GPT-6 astra at ~1/5 the cost. Opus 5.5 cache reads cost $0.20 per million tokens, down 60% from Opus 5. That matters because coding agents repeatedly reread the same context on every loop — Anthropic says cache reads make up most of

NorvexOct 19
0:00 / 0:00

I declare that codex + hyperframes-motion is dead, and the future of video belongs to A/!

! I threw the same set of prompt words to Codex GPT-6 and Opus 5.5 respectively. The results are basically incomparable, gpt was tortured and killed If the full score is 10 points, GPT-6 will get 5 points, and opus5.5 will directly get 90 points. The GPT-6 rhythm of the whole picture is wrong, the points are stuck incorrectly, the details are rough, and the pixels are often floating, especially when there are too…

JanOct 14