Replying to @TonalWeaver @mattpocockuk
Obviously. But at work, I average 20 PRs a day (at Anthropic, they're at 30 per dev per day). We're currently experimenting with different approaches to formalization, with agents delivering up to 100 PRs an hour. That's not something anyone has time to review anymore, even if the PRs are modest in size. Good thing, though, that small PRs can be validated and formalized pretty nicely.
Grokbot, or whoever does the news on this site, gives a pretty good summary of where all this is headed.
Replying to @dummerspast39
At least google put in some effort to benchmaxx their flash models so they at least look good. Terra 6.1 looks giga ass even in their own release blog 😭
Like what is this even.
And maybe it is a decent RAG driver or whatever but this is the last model they need ion a world in which people switching subs for Opus and Fable55 coming soon.
Replying to @kimmonismus
TO DELIVER FAST AND CAPABLE COMPUTER USE AT A LOWER PRICE.
but pls ignore our own benchmak
This has to be a joke right? lol
release blog: openai.com/index/introducing…












