Google just made gemini 3.8 flash much harder to ignore.
It doesn’t just generate the work anymore. It builds, tests, catches its own mistakes, fixes them, then runs the loop again. Why this matters: → Google shipped Gemini 3.8 Flash just 3 weeks after 3.7 Flash. → It’s
Got even better results with gpt 5.6 sol + claude opus 5.
The tamagotchi itself is a high quality image, not an svg. the buttons are interactive, they have shadows, and the whole thing feels much more like a real device i am really curious how much better astra is, but even
ChatGPT-6 Astra is live on RamenBench!
It worked on High for 7min 52sec, and used up 8% of my 5hr session limits on the Plus plan (surprisingly low burn!) As you can see, graphics overall were a HUGE improvement. Astra is also right-handed now. It lost the "get more noodles
Tried this prompt with Grok 4.6, Fable 5.1 and Opus 5.
Nothing gave me an output good enough that I felt like posting it. GPT-6 Astra gave me this. Decent and well composed. One continuous camera move from inside a single GPU die out to a datacenter campus, six stops, the year
You have reached the end of the archive
All of Claude Opus 5