Josh

Just trying to get the hang of this thing

[Josh @JustJorshin]

@thsottiaux @blueemi99 Wait a sec, this makes sense now

[Josh @JustJorshin]

@thsottiaux Love these

But hold up a few hours, I’m cooking real quick

[Josh @JustJorshin]

@theo I will say, never been a fan of Anthropic but opus 5 and fable 5 have got me using them more than none

[Josh @JustJorshin]

@thsottiaux Go ahead and like this for the 750 tp/s 5.6 Sol

[Josh @JustJorshin]

@thsottiaux @Angaisb_ Obvious posting

[Josh @JustJorshin]

@0interestrates Did the final version still feel natural or did it get too diluted?

[Josh @JustJorshin]

@samsantosb The cost difference vs Opus while staying close in quality is hard to ignore.

[Josh @JustJorshin]

Recently grok feels like the only AI that doesn’t make me wait around while I’m trying to think.
That speed alone changed how much I actually use it day to day.
Anyone else notice the same thing?

[Josh @JustJorshin]

@doganuraldesign This is the exact kind of prompt that separates usable generations from stiff ones. Saving it.

[Josh @JustJorshin]

@rauchg Sol still leads absolute scores, but this ratio is what actually gets a model into real workflows.

[Josh @JustJorshin]

@dbatura @XMoney Since I followed X money, can I get it today?

[Josh @JustJorshin]

@sama Love Sol but we need more

[Josh @JustJorshin]

How do you guys feel about having an in-line camera inside of your app? @raroque did an amazing job, implementing it inside of his app and I want to see if it will be the next big thing.

[Josh @JustJorshin]

Grok Hack
Paste your chaotic brain dump and ask “Turn this into a prioritized 3 task list with the easiest 10 min win first.” Instant clarity, zero overthinking.

What’s your go to Grok prompt that actually sticks?

[Josh @JustJorshin]

@paolozamparutti @mark_k Waiting on 4.6/4.7 too. The Build improvements with 4.5 already changed the game.

[Josh @JustJorshin]

@sandiegocausa Love this team approach. Running models against each other is the new pair programming.

[Josh @JustJorshin]

Who wants to give me some hardware to run this beast?

Obviously I will let you use the model after I host it.

This is a great deal if you ask me

[Kimi.ai @Kimi_Moonshot]

Releasing the model weights and technical report of Kimi K3.

Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window.

New model architecture: 2.5x the intelligence per unit of compute, not just more params.

Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale.

Model weights: https://t.co/7m7eEg6Y0B
Tech report: https://t.co/yeu6cjpMCT
Tech blog: https://t.co/YTfiMSNM1f

[Josh @JustJorshin]

@thsottiaux @theo Please be what I think this is

[Josh @JustJorshin]

This is why I run /deep-research before I start outlining any longer piece. Getting the verified claims + clear gaps first stops me from building on shaky sources later. The Partial flag is actually helpful, it tells me exactly where I still need to dig.

Have you paired it with regular Grok chat for the writing part after?

[Josh @JustJorshin]

For clean faceswaps in Grok I’ve had better luck being super specific: ‘Replace only the face with [exact description or clear reference], keep exact pose, lighting, expression, skin texture and background completely unchanged. Photorealistic match, identity preserved, no artistic changes.’ Upload both images clearly. Sometimes adding ‘no artifacts, match lighting precisely’ helps. What references are you feeding it?