Wish we had today's technology when I was a kid with all those figures. Had boxes and boxes and was so pissed of nothing ever was like commercials but still love TMNT. Now you just just aim your phone and make your cartoon practically. Course they dont make toys/entertainment as good now days. Weird.
I thought meta muse spark 1.2 was a good model 3x's time it has taken me on hour plus fixing it's mess because it ignored explicit instructions not to do something and did it anyway ruining working code. Models that you tell do do a b c and completely ignore a, b, and c shouldn't even be beta that is nonsense.
I am really enjoying the @MetaforDevs muse Spark 1.2 model. I had a issue with it pulling a ultra stupid move but it resolved it. It assumed I meant something else when I gave it link to github stack pr merge and started merging them one by one then when I told to read the actual link, things got weird. It ended - resolved though.
@NickADobos@thsottiaux I believe he is saying your friends on their computer are no different than the agents on the remote computer. Did have to read it twice.
Kinda feel bad for the codex meta spark agent thing. Codex+Sol got to handle 8 PRs over hours. He got the shit-end of the stick merging all it's lose-ends because ran out of tokens not realizing the "reset" means you next reset is pushed out 7 days.
Since I been using the harness as agent identities codex security has found 18 security bugs authored by codex and 22 authored by claude. Not apples to apples but interesting.