Replying to @END_OF_PAIOTU
ℤ/6ℤは体ではないため、風邪をひきません。
perhaps a better translation?
I don't need to replace Google, I want my models to actually be smart to be able to do things themselves. I want them to be able to make useful things, reliability.
I was annoyed by Gemini, tool calls being completely opaque is very frustrating, but having fewer tools means
I subscribed to T3 chat (of @theo fame) and it's much worse than I expected. The extra performance is nice over the complete performance disaster that is Gemini, and the more clarity on tool calls is also appreciated, but the fact it's missing so many features, such as canvas,...
is a deal breaker. Both for text and for code, canvas allows you to iterate on the same material repeatedly, which is a huge win, and having AI code is often much better than having it attempt to do all reasoning itself.
Replying to @OfficialLoganK
@OfficialLoganK please fix the rate limit situation for Gemini, manually implementing back off is a pain and it doesn't help that there's for some reason a rate limit for a data center that's lower than the overall rate limit
price transparency would also be great! if you could
Why doesn't Codex CLI work with o4-mini? @OpenAIDevs It does work if I select a different model, like gpt-4.1
"buffer amplifier" from Gemini 2 Flash multimodal (I tried other models but they were even worse)
This is completely wrong for at least 8 reasons
ash retweeted
One does not simply become a master of karate. First, you must accidentally walk into a spider web.
Oh my god, I am literally in tears right now. Biden is pausing approvals and directing the DOE to expand its evaluation of new liquid methane gas terminals to consider their impact on #ClimateChange. Thank god, thank god.











