Love your work, Ruben. You helped me set up Claude for my mental health/addictions consultancy practice. I do plenty of pro bono work too. Quick question: when you say ‘Grok for search, Claude for execution’, how do you run this in practice? Just copy and paste your Grok results over to Claude? Stay safe over there, brother. Thanks
I really love to hear it, Rupert. In practice, I ask Grok for a summary of all of the key findings that he made during the research, and then I upload everything to Claude.
I am reconfiguring my tech stack and will take at Grok again. I am always optimizing my workflow. Perplexity as an aggregator used to have Grok but I don't see it on the list anymore. Appreciate your diligence and hope you are safe.
Same, Grok has been the best so far. Used to be Perplexity, but Grok just replaced all of my Perplexity queries. I only trust it now. I feel like Perplexity is lagging behind. Maybe their computer super orchestrator move will be the right one, but so far it's $200 a month, so I don't wanna try; it's just too expensive.
And I would not have tried Perplexity Max, if it had not been for a computer glitch on their part that sort of "prompted" me to keep building out my workflow there. That said, and I don't want to admit it given my latest issue with them, but it will likely pay for itself.
Also you build an at entirely different level than I do so, there is that.
As someone who is at the beginning stages of learning/researching A.I., I'm as interested in what's available in free models as in the paid versions. Understanding that technology shifts rapidly in this space, has the testing you've done identified which free models are best at this task versus requiring a paid subscription?
Great content, as always, filled with insights for both the prompts and the results.
You really need to pay for AI. The difference between the free and the paid version is just so big that there is no way in hell you're not going to make your money back with it. Just paying for AI tools puts you ahead of 99% of the population, and then knowing how to use them puts you ahead again, probably 80% of the people who are paying in the first place. Just too much to be done, and these tools do cost a lot of money, I know.
Such an interesting post thank you for always doing your diligence and being so well researched!
I've recently been experimenting with Grok in my own research and I've consistently been impressed. I really think it's one of the best models out there but a lot of people are reasonably scared away because of who owns it.
Have you done anything with it's image generators? That's the thing that totally blew me aways a few weeks ago.
I don't trust anything but Grok these days, which has been my staple. There's something about trusting LLMs and just using the right one every time you need it. I do test ChatGPT from time to time because they surprised me with extended thinking recently.
Great article, Ruben. Routing tasks through each model's competitive strengths builds quite a powerful search tool. How would you think about structuring this kind of capability on OpenRouter?
I don't only hate Grok because I hate Elmo. I hate Grok because it's allowing boys to generate and circulate fake nudes of their classmates and traumatizing minor girls and the people who love them. There's even a class action lawsuit now.
That said, I don't plan to unsubscribe. I'm happy to read your other suggestions and advice.
This is a great breakdown. I’ve been testing AI from a slightly different angle, not just which tool is best for search, but how to actually turn it into a repeatable content system. Last weekend I used AI to build a full week of content in less than 2 hours across platforms. It really made me rethink things. The shift isn’t just better search, it’s what you do with it after.
Yes! Perplexity is Step 4 in my weekly system. Monday I plan with Claude. Tuesday I write the anchor piece. Wednesday I repurpose. Thursday I use Perplexity to research topics for next week and bank ideas in Notion. Search isn’t where I start. It’s where I refuel. After reading this, I’m curious if Grok changes that Thursday step. 272 sources in 37 seconds is hard to ignore ;)
You can try Supergrok 3 days for free before the 30$
This is great news, Nelly. Thank you for sharing with everyone else.
it’s really good 2. the beta for 4.20 with the 4 agents works so well
Recently it was a 7 days. Why did Elon reduce trial?
Ask Grok 🤣
Love your work, Ruben. You helped me set up Claude for my mental health/addictions consultancy practice. I do plenty of pro bono work too. Quick question: when you say ‘Grok for search, Claude for execution’, how do you run this in practice? Just copy and paste your Grok results over to Claude? Stay safe over there, brother. Thanks
I really love to hear it, Rupert. In practice, I ask Grok for a summary of all of the key findings that he made during the research, and then I upload everything to Claude.
Ok cool, that is the solution I stumbled upon today. Thanks for the reply 👍🏼
You’re welcome!
I am reconfiguring my tech stack and will take at Grok again. I am always optimizing my workflow. Perplexity as an aggregator used to have Grok but I don't see it on the list anymore. Appreciate your diligence and hope you are safe.
Same, Grok has been the best so far. Used to be Perplexity, but Grok just replaced all of my Perplexity queries. I only trust it now. I feel like Perplexity is lagging behind. Maybe their computer super orchestrator move will be the right one, but so far it's $200 a month, so I don't wanna try; it's just too expensive.
And I would not have tried Perplexity Max, if it had not been for a computer glitch on their part that sort of "prompted" me to keep building out my workflow there. That said, and I don't want to admit it given my latest issue with them, but it will likely pay for itself.
Also you build an at entirely different level than I do so, there is that.
If it pays for itself, that's really all that matters.
What a monster of a post! Thank you for the work that went in to this.
twice a week Wednesday and Sunday. Stay tuned
Really well done. Thank you 🙏
Thanks, John.
As someone who is at the beginning stages of learning/researching A.I., I'm as interested in what's available in free models as in the paid versions. Understanding that technology shifts rapidly in this space, has the testing you've done identified which free models are best at this task versus requiring a paid subscription?
Great content, as always, filled with insights for both the prompts and the results.
You really need to pay for AI. The difference between the free and the paid version is just so big that there is no way in hell you're not going to make your money back with it. Just paying for AI tools puts you ahead of 99% of the population, and then knowing how to use them puts you ahead again, probably 80% of the people who are paying in the first place. Just too much to be done, and these tools do cost a lot of money, I know.
absolutely have been having a blast with the 4 agents here. watching each one act differently based on their context and system prompt is really fun.
Now try specializing them for your exact job :)
Great article on Search. Thank you and keep up the good work.
Thanks, John. I'll try to keep up the standard just as high.
Ruben, you are a flat out legend
thanks man :))) means a lot
Such an interesting post thank you for always doing your diligence and being so well researched!
I've recently been experimenting with Grok in my own research and I've consistently been impressed. I really think it's one of the best models out there but a lot of people are reasonably scared away because of who owns it.
Have you done anything with it's image generators? That's the thing that totally blew me aways a few weeks ago.
I don't trust anything but Grok these days, which has been my staple. There's something about trusting LLMs and just using the right one every time you need it. I do test ChatGPT from time to time because they surprised me with extended thinking recently.
Great article, Ruben. Routing tasks through each model's competitive strengths builds quite a powerful search tool. How would you think about structuring this kind of capability on OpenRouter?
i prefer using it directly on Grok.
I don't only hate Grok because I hate Elmo. I hate Grok because it's allowing boys to generate and circulate fake nudes of their classmates and traumatizing minor girls and the people who love them. There's even a class action lawsuit now.
That said, I don't plan to unsubscribe. I'm happy to read your other suggestions and advice.
Its search capabilities are genuinely better right now - thats true and what I focus on.
Hey Ruben, Slack link is no longer working for me (I did not used when you send it). Please resend the link to Slack again
DMed you an updated one :)
Love the art work!!
appreciate it, Dave!!
Love the art work!!
thank you :) can you guess from what tool?
Nano Banana 2?
This is a great breakdown. I’ve been testing AI from a slightly different angle, not just which tool is best for search, but how to actually turn it into a repeatable content system. Last weekend I used AI to build a full week of content in less than 2 hours across platforms. It really made me rethink things. The shift isn’t just better search, it’s what you do with it after.
thanks for sharing!! :) is search a part of system?
Yes! Perplexity is Step 4 in my weekly system. Monday I plan with Claude. Tuesday I write the anchor piece. Wednesday I repurpose. Thursday I use Perplexity to research topics for next week and bank ideas in Notion. Search isn’t where I start. It’s where I refuel. After reading this, I’m curious if Grok changes that Thursday step. 272 sources in 37 seconds is hard to ignore ;)
I wrote about the system I tested here if you’re curious:
https://maramartins.substack.com/p/create-week-of-content-2-hours-ai?r=48946w&utm_medium=ios