Grok 5 Could be xAI's Biggest Breakthrough Yet...

Reminds me of the old days on Stack Overflow... you would ask a question, and some asshat developer who thought he knew everything, would not answer your question, but instead question why you would even want to know that. Example: "Hey I have a million files I need to delete, what is the fastest way to do that?" Asshat answer: "I dont know why you would even be in that situation, you should look into why you have so many files in the first place and enact new policies to ensure that un-needed files are not created in the first place." Uh huh, just tell me how to delete the fcking files.

I just uploaded a selfie I took and let Grok roast me in unhinged mode mode. Holy shit, it was savage, and honestly made a few good points. I'm not sure I'll ever leave the house again lol

They better bring the spicy mode back unnuetered!

Grok 4.20 edited some JavaScript app I have, perfectly. After the previous thinking Grok model produced garbage edits. So I'm hopeful xAI is catching up.

Try Caffeine for free: https://cffn.ai/WesRoth If you've ever had an app idea sitting in the back of your head, this is your sign to go and build it.

I like that it shows me what’s wrong with my thinking. I use a few LLMs and this one is the most important one for me, because it saves me from making errors that the other models didn’t see or didn’t flag

Breakthroughs by day, Wes Roth Podcast by night, all day.

Thanks Wes, keep 'em coming...

Great video 🙌 I use Claude a lot and recently tried something called context-mode. It basically prevents huge tool outputs from flooding the context (saved ~18k tokens in one run). Feels like a missing piece for heavy Claude users. Would be interesting to see this kind of stuff covered.

Wes, thanks for the video. Can we rule out the likelihood of the contrarian reflex being topic-dependent?