Nick

Serial HODLer $TRENCH

Quitting is the real failure. | WAGMI

[Nick @saintXsol]

Retweets ≠ Endorsements. Just here for the memes and the money. Don’t take it personally—I probably don’t either.

[Nick @saintXsol]

@R2Cdev_ Surely not for anything ‘agentic’

[Nick @saintXsol]

@rohit3a It's not that xhigh sucks, it's just that max is more thorough and produces higher quality output.

It's really noticeable on open ended tasks.

They all get the job done, but one put in more care into every step.

To see the difference, ask them to create a video, then compare.

[Nick @saintXsol]

@rohit3a They all have thier use cases. I mostly use high and medium, but there are some tasks I give to max and it leaves me speechless (esp. compared to multiple high/xhigh)

[Nick @saintXsol]

@uzairakrum Oh wow, that's insane.

[Nick @saintXsol]

@ycombinator @demishassabis @garrytan @elonmusk please can I get the freaking YouTube Link!!!

[Nick @saintXsol]

@lyraxana Brave of us to assume we're getting access. It would probably be usage credits only

[Nick @saintXsol]

@theo This is just propaganda and borderline retarded

[Nick @saintXsol]

@robinebers Don’t fall for the trap 😭
You’re basically donating $5 to Meta

[Nick @saintXsol]

@theo Well, most people are using Muse Wonder not “Full”. I can’t speak for Full, but the regular Womder is benchmaxxed

[Nick @saintXsol]

@petergostev Nonstop OpenAI glazing from you lately

[Nick @saintXsol]

@Saboo_Shubham_ The Foresight app behaves like malware… it’s too invasive.

[Nick @saintXsol]

@scaling01 Let’s see if you’re still saying the same thing tomorrow.

Who am I kidding, you probably would just accuse them of benchmaxxing, just o you can say you’re right

[Nick @saintXsol]

@AIatMeta @TheINAOG Drop the model size for both or you’re lying about opening the weights

[Nick @saintXsol]

@thtbee_ @lyraxana tomorrow... relax.

[Nick @saintXsol]

@theo I agree with you that people discount models because of 3D stuff, but Muse Wonder is also somewhat benchmaxxed.

I tested it on an ood prompt (python) and it performs worse than Sonnet & 6.1 Sol. Tested all three on high effort: Sonnet > Sol > Muse Wonder

[Nick @saintXsol]

@AdamHoltererer Or maybe just bad at 3D

[Nick @saintXsol]

@scaling01 Ngl, that’s a sign of benchmaxing