Tomek W

[Tomek W @QFallenq]

@AIMind_Ai Not really possible..
What did you use?

[Tomek W @QFallenq]

@theo max reasoning is only for niche use cases.. Like inventing new technologies..

[Tomek W @QFallenq]

@maxxrubin_ @AdamHoltererer Probably benchmaxxed..

[Tomek W @QFallenq]

@theo Max can be used rarely for scientific stuff..

[Tomek W @QFallenq]

@aaravdayal @Panchoavila @barundebnath You forgot the x... Unless you meant dollars.. Or seconds.. Idk

[Tomek W @QFallenq]

@aaravdayal @Panchoavila @barundebnath It was 450x obviously, idk what you're talking about

[Tomek W @QFallenq]

@Lummox_eth Are we testing harness or what?? Your test doesn't make any sense, you didn't even say what model doesc claude code use, and what model is using the blender mcp..

[Tomek W @QFallenq]

@synthwavedd One thing, cracking games..

[Tomek W @QFallenq]

@EnvolDev There's more detail in some places, less in other, for example, on the launch version of opus 5.5 the train phases thru a part of the bridge, because well, that said part of the bridge is put on the rails.. The background scene on today's example is also better..

[Tomek W @QFallenq]

@EnvolDev Looks better now

[Tomek W @QFallenq]

@notjazii It is routed in 3 out of 10 convos..

[Tomek W @QFallenq]

@SadAlbert10 and it's probably opus 5.6

[Tomek W @QFallenq]

@SadAlbert10 In 50% of convos it is routed in 50% it's not, if it says it vaguely knows it it's routed

[Tomek W @QFallenq]

@weswinder Honestly it's quite useful for me... Just not astra..
Fable 5.1 on max effort is the way to go for scientific research and such.. Opus 5.5 is extremely close but there's still a gap.. I think it's the only place where opus 5.5 loses agains fable..

[Tomek W @QFallenq]

@goon_nguyen It's still theoretically coding, but yeah..

[Tomek W @QFallenq]

@thsottiaux .com

[Tomek W @QFallenq]

@wholyv It's better at modeling guns and stuff..
Opus holds back because of "safety reasons", look at it's reasoning

[Tomek W @QFallenq]

@theo Opus Has well developed network thinking