I cancelled Grok Heavy today. I had found myself questioning Grok more and more in recent weeks and using the lesser versions instead of Heavy because 90% of the time Heavy would fail (time) out after 5-10 minutes of running its agents on a query.
But those weren’t the reasons why I cancelled.
I loaded up GeminiPro this morning. Subscribed to its Grok Heavy equivalent and started to use it. High level: GeminiPro outclasses Grok Heavy in every conceivable way. From reliability to function to feature set to speed.
This is hard for me to say because (1) I am an Elon fanboy and (2) I despise Alphabet. Alas, you have to give the devil his due.
So why the shift? Today, 3 distinct reasons:
(1) I saw a simple opportunity for a test question for Gemini, 12 minutes before the close today. I was pretty confident as to the why we saw a spike 12 minutes to close so I wanted to see what Gemini would think.
Less good than Gemini but better than the atrocity it put forth as facts as 4.1 (thinking).
(2) My grandmother is 100 years old. About a year ago my mother asked me to try and restore this photo taken of her a long, long time ago. The photo, as you will see, is well beyond the capability of anything that exists currently. Until today. I uploaded the photo to Gemini, then I uploaded 3 reference photos of my grandmother. One from when she was a teenager. Two from more recently. I then asked Gemini to try and fix the photo from the provided reference. Here’s what it did. You can see how thrashed the old photo was. Now I should note that it’s not a perfect recreation because I only had 3 pictures on hand. I plan on supplying Gemini with many more to see what results I can get, but I am confident given the initial test that it will actually recreate the picture. Mind blown. Grok told me about the picture like it was a live action roleplay. Heavy failed to even try.
(3) My dad died a long time. Late 90s. As those that have lost ones in the pre-iPhone era know, pictures are often very difficult to come by. They’re all over the place, and in a family the size of mine, who even knows.
Anyway we have this old video of him returning from a business trip from Amsterdam and we always thought it was a beautiful picture of him and my mom reunited after not seeing each other for 2 years. The problem was it was a screenshot from a VHS video.
Now I have tried to fix this picture many times. Including with Grok; Fotor, Photoshop, you name it. Always with terrible results.
Anyway I decided to give it Gemini the test. I had low expectations as nobody had been able to fix this picture to my liking.
Lo and behold. First try. Home run. I literally cried. I couldn’t believe it. It had made my dad look more like my dad than the grainy old video still.
So in the interest of science i put the exact same prompt into Grok Heavy. Grok 4.1 (thinking). Maybe it had gotten better.
Warning, it had not gotten better. What follows is a real interaction in real time with Grok Heavy to upres an old pic. I really just want to drive home the point that this is not a joke. This is what Grok Heavy thought was appropriate.
Gm sir, I actually don't see a big difference ... both AI platforms say similar things... in a nutshell a big seller sold it. I am in EST time is 12:45 PST 3:45 EST?
The good side of Grok is it provides thirty X posts and 10 web pages as sources. Both will disclaim limited paywall access which is really the meat in the sandwich. I like the idea of getting charts which grok sux at so I use Chat GBT if I want them but I have CQG and Schwab live. I ldo check stuff from AI!! Not sure what the competitive line is for image generation. I'm not sure I want to know!
I doubt I will ever cross that AI line. I prefer my memories and imagination. What I have found from computer analysis for the last 20 years is the computer/ AI will strive to give you the answer you ask for... a rather inferior iteration or identity of your own mind often leading frustrating losses in trading. So, I stopped it years ago and never went back.
Usually I proof and correct my writing myself which is why I make so many mistakes! But unfortunately mistake in an author's writing have become a badge of honor... If I could afford an editor I'd pay one and write a lot more. I do use grok as a souped up spell check if I have been writing quickly sloppy messy stream of conciousness and I'm late.
In the end these are just machines like any other mechanical convenience. I once almost cut my leg off with a 20 inch chain saw bc I was working all day and I was exhausted. An instructive lesson. You are such a good writer. If I may respectfully suggest... Draw line between that and AI before it consumes you.
I used to write a lot. Professionally. I’d often tell people that I’ve been on many best sellers lists, but only as the “Bridesmaid.” Anyhow it’s past me and not something I’m ready to engage with again at any real “competitive” level.
Imagine generation as well as speed and accuracy of information scraping, aggregation and distillation is more a test of the performance of the AI. In that use case, putting my $300/month Grok vs my 125/month Gemini (249/month after first 3 months) was a no brainer. I am unable to even use Grok heavy anymore because 90% of its queries fail out.
I would gladly go back to Grok; but right now I have to pull teeth and fact check him on financial numbers because they are almost always all over the place.
I still have to put Gemini through its paces but currently this stands atop the leaderboard for me.
good information. I will immediately commence using Gemini. And I am a double Gemimi so there is that, too. Good stuff sir. PS I wish ... we didn't have it but like the river we can't step into twice, once we are in it we can't go back.
Of course you know ... since I enable gemini ... Gemini is reading our onversations right now? I just checked and Gemini quoted ME saying, " I am a double Gemini." in the previous reply to you.
I remember this line from 50 years ago "I could drink a case of you and still be on my feet." I never knew what she meant, but it was Joni, and you were never sure of what she meant, as gifted a communicator as she was. Nice pick JJ, as it relates to markets now.
I promise this is relevant.
I cancelled Grok Heavy today. I had found myself questioning Grok more and more in recent weeks and using the lesser versions instead of Heavy because 90% of the time Heavy would fail (time) out after 5-10 minutes of running its agents on a query.
But those weren’t the reasons why I cancelled.
I loaded up GeminiPro this morning. Subscribed to its Grok Heavy equivalent and started to use it. High level: GeminiPro outclasses Grok Heavy in every conceivable way. From reliability to function to feature set to speed.
This is hard for me to say because (1) I am an Elon fanboy and (2) I despise Alphabet. Alas, you have to give the devil his due.
So why the shift? Today, 3 distinct reasons:
(1) I saw a simple opportunity for a test question for Gemini, 12 minutes before the close today. I was pretty confident as to the why we saw a spike 12 minutes to close so I wanted to see what Gemini would think.
https://g.co/gemini/share/194903e546f7
Correct.
“Groks 4.1 (Thinking)” answer to the same prompt?
https://grok.com/share/bGVnYWN5LWNvcHk_6bbd64b3-90d0-443c-a766-5d6075f9d462
Wrong. Down to getting timezone wrong. 5 hours off the mark. Just a fairy tale.
Grok Heavy answer (after 15 minutes mind you)
https://grok.com/share/bGVnYWN5LWNvcHk_7c32a0c9-d7cd-4c0b-acbf-4eae7453f295
Less good than Gemini but better than the atrocity it put forth as facts as 4.1 (thinking).
(2) My grandmother is 100 years old. About a year ago my mother asked me to try and restore this photo taken of her a long, long time ago. The photo, as you will see, is well beyond the capability of anything that exists currently. Until today. I uploaded the photo to Gemini, then I uploaded 3 reference photos of my grandmother. One from when she was a teenager. Two from more recently. I then asked Gemini to try and fix the photo from the provided reference. Here’s what it did. You can see how thrashed the old photo was. Now I should note that it’s not a perfect recreation because I only had 3 pictures on hand. I plan on supplying Gemini with many more to see what results I can get, but I am confident given the initial test that it will actually recreate the picture. Mind blown. Grok told me about the picture like it was a live action roleplay. Heavy failed to even try.
Here is the result:
https://imgur.com/a/wXxoU5x
(3) My dad died a long time. Late 90s. As those that have lost ones in the pre-iPhone era know, pictures are often very difficult to come by. They’re all over the place, and in a family the size of mine, who even knows.
Anyway we have this old video of him returning from a business trip from Amsterdam and we always thought it was a beautiful picture of him and my mom reunited after not seeing each other for 2 years. The problem was it was a screenshot from a VHS video.
Now I have tried to fix this picture many times. Including with Grok; Fotor, Photoshop, you name it. Always with terrible results.
Anyway I decided to give it Gemini the test. I had low expectations as nobody had been able to fix this picture to my liking.
https://imgur.com/a/VviKUV9
Lo and behold. First try. Home run. I literally cried. I couldn’t believe it. It had made my dad look more like my dad than the grainy old video still.
So in the interest of science i put the exact same prompt into Grok Heavy. Grok 4.1 (thinking). Maybe it had gotten better.
Warning, it had not gotten better. What follows is a real interaction in real time with Grok Heavy to upres an old pic. I really just want to drive home the point that this is not a joke. This is what Grok Heavy thought was appropriate.
https://imgur.com/a/a1ZjVdZ
I wish I was joking. I stared at this picture for a half an hour. Laughing and crying because it was so outlandish.
And that is how I realized we’re now left with Gemini, OpenAI and a handful of other serious AI competitors and a lot of absolute -certain- failures.
Gm sir, I actually don't see a big difference ... both AI platforms say similar things... in a nutshell a big seller sold it. I am in EST time is 12:45 PST 3:45 EST?
The good side of Grok is it provides thirty X posts and 10 web pages as sources. Both will disclaim limited paywall access which is really the meat in the sandwich. I like the idea of getting charts which grok sux at so I use Chat GBT if I want them but I have CQG and Schwab live. I ldo check stuff from AI!! Not sure what the competitive line is for image generation. I'm not sure I want to know!
I doubt I will ever cross that AI line. I prefer my memories and imagination. What I have found from computer analysis for the last 20 years is the computer/ AI will strive to give you the answer you ask for... a rather inferior iteration or identity of your own mind often leading frustrating losses in trading. So, I stopped it years ago and never went back.
Usually I proof and correct my writing myself which is why I make so many mistakes! But unfortunately mistake in an author's writing have become a badge of honor... If I could afford an editor I'd pay one and write a lot more. I do use grok as a souped up spell check if I have been writing quickly sloppy messy stream of conciousness and I'm late.
In the end these are just machines like any other mechanical convenience. I once almost cut my leg off with a 20 inch chain saw bc I was working all day and I was exhausted. An instructive lesson. You are such a good writer. If I may respectfully suggest... Draw line between that and AI before it consumes you.
I used to write a lot. Professionally. I’d often tell people that I’ve been on many best sellers lists, but only as the “Bridesmaid.” Anyhow it’s past me and not something I’m ready to engage with again at any real “competitive” level.
Imagine generation as well as speed and accuracy of information scraping, aggregation and distillation is more a test of the performance of the AI. In that use case, putting my $300/month Grok vs my 125/month Gemini (249/month after first 3 months) was a no brainer. I am unable to even use Grok heavy anymore because 90% of its queries fail out.
I would gladly go back to Grok; but right now I have to pull teeth and fact check him on financial numbers because they are almost always all over the place.
I still have to put Gemini through its paces but currently this stands atop the leaderboard for me.
good information. I will immediately commence using Gemini. And I am a double Gemimi so there is that, too. Good stuff sir. PS I wish ... we didn't have it but like the river we can't step into twice, once we are in it we can't go back.
Thus Spoke Zarathustra
Of course you know ... since I enable gemini ... Gemini is reading our onversations right now? I just checked and Gemini quoted ME saying, " I am a double Gemini." in the previous reply to you.
Are you not entertained?!
I remember this line from 50 years ago "I could drink a case of you and still be on my feet." I never knew what she meant, but it was Joni, and you were never sure of what she meant, as gifted a communicator as she was. Nice pick JJ, as it relates to markets now.
Really enjoyed this one, The quote, The art and the vibe really ran hard. Thanks JJ.
“ It’s like giving your kids swimming lessons in a shark tank. It’s also obvious the “jobs” are the “lambs.”
Superb!!