The company that kicked off this "revolution" seems to be in a spot of trouble. Their last model was a disappointment according to reviews. Google's latest Gemini 3 model is supposed to be vastly superior.

He seems to be doing the right thing of re-focusing his team so I am curious to see where OpenAI ends up. As far as for my personal work, I always use ChatGPT but I will soon explore Gemini again. It wasn't very good a year ago. Grok is also decent.

OpenAI CEO Sam Altman declares 'code red' as ChatGPT competition mounts
https://finance.yahoo.com/news/openai-ceo-sam-altman-declares-code-red-as-chatgpt-competition-mounts-141751345.html

Quote:

---

OpenAI (OPAI.PVT) CEO Sam Altman has issued an urgent memo to employees to accelerate improvements to ChatGPT as competition intensifies from rival AI developers Google (GOOG, GOOG) and Anthropic (ANTH.PVT).

Altman told employees in an internal memo on Monday that he was declaring a "code red" to dedicate resources toward bettering ChatGPT, given the pressure from rivals, The Information reported. The move will delay OpenAI's work of introducing other products, such as AI agents, the outlet said.

Google introduced its latest AI model, Gemini 3, in November, to widespread praise. The debut sent the stock soaring to a record high as Gemini 3 beat ChatGPT on benchmark tests that score AI models' performance.

Salesforce CEO Mark Benioff was among the most prominent tech boosters to weigh in on the release, saying he was ditching ChatGPT for Google's latest model. Altman congratulated Google publicly on social media, but the CEO privately told OpenAI employees that Gemini 3 could create economic headwinds for OpenAI, as reported previously by The Information

---

I'm not an OpenAI fan, but I thought the latest chatgpt has been better than the latest grok.

How do you guys measure this?

Im guessing by click bait article clicks. So we all lose.

I have Gemini and ChatGPT write independent scripts and I'll check their quality. To me Claude generates the highest quality tools, but the token usage is absurd to the point where it isn't useful

OpenAI has no moat other than name brand recognition. google's models are just as good if not better and they have the advantage of being able to integrate it into every product they have. plus they actually make money.

Gemini absolutely sucks.

BTW, I watched a vid the other day where ChatGPT was used by a shop to do successful diagnosis on some broken vehicles. I love my AI future!

Philip J Fry said:

---

I have Gemini and ChatGPT write independent scripts and I'll check their quality. To me Claude generates the highest quality tools, but the token usage is absurd to the point where it isn't useful

---

What kind of tools?
I have not used Claude yet but I am told it is for code gen.

Old McDonald said:

---

OpenAI has no moat other than name brand recognition. google's models are just as good if not better and they have the advantage of being able to integrate it into every product they have. plus they actually make money.

---

Exactly.
They have the ecosystem of GCP that makes it easy to use. With OpenAI, they have to do an integration and there are hassles. The model has to be really amazing to get past Google

Logos Stick said:

---

Gemini absolutely sucks.

BTW, I watched a vid the other day where ChatGPT was used by a shop to do successful diagnosis on some broken vehicles. I love my AI future!

---

Gemini 3 also?
The earlier Gemini models were woke that would happily make jokes on Jesus but scold me if I asked it to make a joke on Muhammad.


I am told Gemini 3 is good but haven't tried.

gemini 3 is world class. i finally switched over from ChatGPT and it's way better.

google is eating Sam's lunch. they're the one fully integrated AI company (own their own silicon, compute, models, and products) and were leagues ahead of OAI on all fronts but the models until gemini 3 came out.

i declined a job opportunity from them just last month for those reasons. bearish and uncertain future.

Grok is my favorite thus far

Logos Stick said:

---

Gemini absolutely sucks.

BTW, I watched a vid the other day where ChatGPT was used by a shop to do successful diagnosis on some broken vehicles. I love my AI future!

---

I could do that 10-years ago with google and message boards. I guess ChatGPT was 5-minutes faster?

Also, I always had to double check the 'instructions' I found online b/c they were not always complete. How does ChatGPT know the data they're pulling from is valid? Unless every auto manu publishes their manuals online (not likely) I don't see how this is better when you still have to be able to confirm the 'fix' is correct.

Of course, use case matters.

Gemini is head and shoulders above any AI that I have tried for image editing. Reasonably close to the better/best stand-alone apps.

Maybe technology has passed me by. I'm not using any AI tools and really don't even know where to start.

A. G. Pennypacker said:

---

Maybe technology has passed me by. I'm not using any AI tools and really don't even know where to start.

---

just let skynet do it's thing. You'll just see a bright flash and then it will be over a millisecond later.

A. G. Pennypacker said:

---

Maybe technology has passed me by. I'm not using any AI tools and really don't even know where to start.

---

Just go to perplexity.ai and enter what you just wrote.

Old McDonald said:

---

gemini 3 is world class. i finally switched over from ChatGPT and it's way better.

google is eating Sam's lunch. they're the one fully integrated AI company (own their own silicon, compute, models, and products) and were leagues ahead of OAI on all fronts but the models until gemini 3 came out.

i declined a job opportunity from them just last month for those reasons. bearish and uncertain future.

---

You declined a job from OpenAI? Wow, congratulations, to get an offer is pretty cool. What field?

I think OpenAI will sell out.

Let me guess they need a government bailout

A. G. Pennypacker said:

---

Maybe technology has passed me by. I'm not using any AI tools and really don't even know where to start.

---

If you are interested, I would start with something you are already doing. I use it mainly for programming. things I am curious about, relearning some electronic/micro-processor stuff.

If you are interested in photography and/or have a bunch of photos -- image editing and/or video generation? I used Grok for my Christmas cards' images last year instead of Adobe Illustrator/ Photoshop.

To get started, tell the AI what you want to do and see if it thinks it can do it, or would recommend a different AI. Then ask for a step-by-step tutorial.

agracer said:

---

Logos Stick said:

---

Gemini absolutely sucks.

BTW, I watched a vid the other day where ChatGPT was used by a shop to do successful diagnosis on some broken vehicles. I love my AI future!

---

I could do that 10-years ago with google and message boards. I guess ChatGPT was 5-minutes faster?

---

You absolutely could not do that 10 years ago. He was interacting with it via voice. It would tell him what to and how to do it and then respond with the next step after he told it the results. It led him through the entire process like having an expert auto mechanic on the phone would do. Doing a generic search that yields 100 solutions - and a whole lot of absolute trash - in no particular order, that you have to wade through and discern is not in the same universe. Voice recognition 10 years ago was also pure crap.

TexAgs91 said:

---

I'm not an OpenAI fan, but I thought the latest chatgpt has been better than the latest grok.

---

I need to add to this. I was referring to coding above, which chatgpt beat grok on.

I asked Grok the CFP rankings after it came out tonight. It was wrong. Then it said it would get it from the source. It was still wrong. It said it apologized and said it tried projecting from the other 2024 rankings (which is ridiculous and it also lied about looking it up).

I asked chatgpt the same question. The first answer was correct.

I thought grok was passing chatgpt up. That is completely false.

I use Copilot and Grok for technical content and coding. I find Grok much better than Copilot for my uses. Haven't used Gemini much in the past year, but probably should give it a go again.

infinity ag said:

---

Old McDonald said:

---

OpenAI has no moat other than name brand recognition. google's models are just as good if not better and they have the advantage of being able to integrate it into every product they have. plus they actually make money.

---

Exactly.
They have the ecosystem of GCP that makes it easy to use. With OpenAI, they have to do an integration and there are hassles. The model has to be really amazing to get past Google

---

GCP?


YEAH YOU KNOW ME!

I tend to use Claude most

I have messed around with most of them. I use Perplexity for most of my daily work automation right now. But got dang this stuff is progressing so fast I have moved over certain tasks between several different ones over the last year or so as they get better.

To echo an above post, it really matters what task you're trying to do.

For example, Perplexity is really good right now doing automated research tasks id pay an analyst to do. And aggregating news/market intel (including scraping social media like LinkedIn) and sending it to me in a very usable way. It also scrapes PDFs and spits out clean data for us exceptionally well.

It runs a "deep research" once a week and sends us new potential buyers to add to our target list and they all have made sense which I find remarkable.

I trained Grok to run one of our most used calculations and scared to try to move that as it's too locked in. I say, "let's go" and it immediately prompts for the inputs it needs or an excel with several runs with said inputs and has been shockingly accurate.

Wish I had more time to Jack with it. We'll have a new fresh guy in May and we were making sure that hire has a background (and interest) in this **** bc it's real and saves us a **** load of time. (It's just me and a partner fwiw that run our service line in a much larger firm but very independent)

I envy those that own a private enterprise and can fully leverage all this and connect everything together. So much competitive edge we leave on the table due to guardrails.

Replit has been super fun to play with but we can't API or connect **** unfortunately and CP for outlook and the rest suck unless you have a data engineer ready to help you along (which know corporations do)

GPT as a product is better than Gemini.

Nothing beats claude code for the price, or even comes close

The speed in which all of the models are improving is crazy. I've been a chat guy for a while and have gotten used to it. Switching would need to be for something that chat can't do for me. It is kinda my word processor.

I do some GRC consulting and use chat to help generate personas to create scenario reactions to build incident response communication plans, table top exercises etc. it's great for that.

Forgive the coming wall of text. I get excited.

Despite the consulting, my passion is being an indie author. I have 4 books and a novella published with about 4500 books sold which is better than 90% of authors but not yet in the black.

This year, I have gone all in on Chat for my research and world building with great results. I'm currently writing a fantasy trilogy, and a rom com novella by typed hand. If life does not interrupt me, I can write a 300 page book, have it edited, and the art work done by my sub contractors, format it, and have the layout done by another contractor in 4-6 months at a cost of ~$5k.

I know authors who have embraced AI a while ago to take their thinly written prose and puff it up, and they are cranking a book every couple of months, because they let a company handle all the publishing tasks. Then there are a ton of authors I suspect pump a book a month or less using AI. If they have an existing audience they usually get high ratings regardless.

So the pressure is on to adopt the tech as a business person, but as a creative, I know I have to keep writing to get better. My prose from book 1 - book 4 has seen a vast improvement, so I will keep typing or dictating for the majority of my projects. At least for now.

Every six months I have used chat to see if it can write to my level and in my style. It just got to that level for me, though maybe it was there 6 months ago. Its generating of a plot has gotten better as well, but it is still very basic. Interesting stories are about interesting characters doing unexpected things and at this point it still falls short.

All that said, my prompting has gotten a ton better so maybe that is the difference in this attempt. Because of the success of this test, I decided to add a human directed and edited, AI written book into my pipeline. I will publish this under a different pen name, and unlike so many writers lying about it, I will let people know what they're getting.

But I will not share that pen name with my self-written audience as AI haters will destroy my ratings on my other books.

I'm going to use it to write stories that are at the bottom of my pipeline and would probably never get written at the speed I write and publish myself. I want this project to be as much AI created as possible, and want to see if I can get the formatting, cover design, etc from AI.

After several attempts, I'm resorting to a process that requires me to iterate chapter to chapter. The AI loses frame of reference on which characters know what, the background of some characters, is wrong on its research usage at times, and loses POV.

The AI does a lot wrong but it can now generate 80% accurate/entertaining prose that I can clean up and embellish in one or two quick passes. I edit a chapter and then feed the book up to that point into the next chapter prompt. It is taking some time as I change up the plot and go back to rework chapters. Sometimes things don't occur to you until you read it.

I have not had 40 hours straight to work on it, but I can envision finishing a 300 page book in 3 weeks if I can focus. Even if I can't do most of the publishing aspects via AI, I think I can publish this book for $500-$1000.

One cool aspect of this is that I'm getting to read the book like one of those pick your own adventures, each chapter is new to me even though I know the major beats.

I will finish this first book and publish it, then I will add an AI day to my writing schedule, 3 on the trilogy, one on the rom com, and one on the next AI book. When I'm done writing the trilogy, we'll see how many books in total I will have to publish.

They all leap frog each other over time. GROK 4 is the best engine for math right now, as an example. Declaring one better than the other and declaring the race over - like Benioff did - is ignorant. I suspect he has a lot of Alphabet stock.

Stressboy said:

---

The speed in which all of the models are improving is crazy. I've been a chat guy for a while and have gotten used to it. Switching would need to be for something that chat can't do for me. It is kinda my word processor.

I do some GRC consulting and use chat to help generate personas to create scenario reactions to build incident response communication plans, table top exercises etc. it's great for that.

Forgive the coming wall of text. I get excited.

Despite the consulting, my passion is being an indie author. I have 4 books and a novella published with about 4500 books sold which is better than 90% of authors but not yet in the black.

This year, I have gone all in on Chat for my research and world building with great results. I'm currently writing a fantasy trilogy, and a rom com novella by typed hand. If life does not interrupt me, I can write a 300 page book, have it edited, and the art work done by my sub contractors, format it, and have the layout done by another contractor in 4-6 months at a cost of ~$5k.

I know authors who have embraced AI a while ago to take their thinly written prose and puff it up, and they are cranking a book every couple of months, because they let a company handle all the publishing tasks. Then there are a ton of authors I suspect pump a book a month or less using AI. If they have an existing audience they usually get high ratings regardless.

So the pressure is on to adopt the tech as a business person, but as a creative, I know I have to keep writing to get better. My prose from book 1 - book 4 has seen a vast improvement, so I will keep typing or dictating for the majority of my projects. At least for now.

Every six months I have used chat to see if it can write to my level and in my style. It just got to that level for me, though maybe it was there 6 months ago. Its generating of a plot has gotten better as well, but it is still very basic. Interesting stories are about interesting characters doing unexpected things and at this point it still falls short.

All that said, my prompting has gotten a ton better so maybe that is the difference in this attempt. Because of the success of this test, I decided to add a human directed and edited, AI written book into my pipeline. I will publish this under a different pen name, and unlike so many writers lying about it, I will let people know what they're getting.

But I will not share that pen name with my self-written audience as AI haters will destroy my ratings on my other books.

I'm going to use it to write stories that are at the bottom of my pipeline and would probably never get written at the speed I write and publish myself. I want this project to be as much AI created as possible, and want to see if I can get the formatting, cover design, etc from AI.

After several attempts, I'm resorting to a process that requires me to iterate chapter to chapter. The AI loses frame of reference on which characters know what, the background of some characters, is wrong on its research usage at times, and loses POV.

The AI does a lot wrong but it can now generate 80% accurate/entertaining prose that I can clean up and embellish in one or two quick passes. I edit a chapter and then feed the book up to that point into the next chapter prompt. It is taking some time as I change up the plot and go back to rework chapters. Sometimes things don't occur to you until you read it.

I have not had 40 hours straight to work on it, but I can envision finishing a 300 page book in 3 weeks if I can focus. Even if I can't do most of the publishing aspects via AI, I think I can publish this book for $500-$1000.

One cool aspect of this is that I'm getting to read the book like one of those pick your own adventures, each chapter is new to me even though I know the major beats.

I will finish this first book and publish it, then I will add an AI day to my writing schedule, 3 on the trilogy, one on the rom com, and one on the next AI book. When I'm done writing the trilogy, we'll see how many books in total I will have to publish.

---

^--- Post written by ChatGPT?

Logos Stick said:

---

They all leap frog each other over time. GROK 4 is the best engine for math right now, as an example. Declaring one better than the other and declaring the race over - like Benioff did - is ignorant. I suspect he has a lot of Alphabet stock.

---

He is a CEO. Remember what CEOs do.
Like you said, he has stock or he is about to make a deal with Google or something where he makes a ton of money cuz he ain't stoooopid.

infinity ag said:

---

Stressboy said:

---

The speed in which all of the models are improving is crazy. I've been a chat guy for a while and have gotten used to it. Switching would need to be for something that chat can't do for me. It is kinda my word processor.

I do some GRC consulting and use chat to help generate personas to create scenario reactions to build incident response communication plans, table top exercises etc. it's great for that.

Forgive the coming wall of text. I get excited.

Despite the consulting, my passion is being an indie author. I have 4 books and a novella published with about 4500 books sold which is better than 90% of authors but not yet in the black.

This year, I have gone all in on Chat for my research and world building with great results. I'm currently writing a fantasy trilogy, and a rom com novella by typed hand. If life does not interrupt me, I can write a 300 page book, have it edited, and the art work done by my sub contractors, format it, and have the layout done by another contractor in 4-6 months at a cost of ~$5k.

I know authors who have embraced AI a while ago to take their thinly written prose and puff it up, and they are cranking a book every couple of months, because they let a company handle all the publishing tasks. Then there are a ton of authors I suspect pump a book a month or less using AI. If they have an existing audience they usually get high ratings regardless.

So the pressure is on to adopt the tech as a business person, but as a creative, I know I have to keep writing to get better. My prose from book 1 - book 4 has seen a vast improvement, so I will keep typing or dictating for the majority of my projects. At least for now.

Every six months I have used chat to see if it can write to my level and in my style. It just got to that level for me, though maybe it was there 6 months ago. Its generating of a plot has gotten better as well, but it is still very basic. Interesting stories are about interesting characters doing unexpected things and at this point it still falls short.

All that said, my prompting has gotten a ton better so maybe that is the difference in this attempt. Because of the success of this test, I decided to add a human directed and edited, AI written book into my pipeline. I will publish this under a different pen name, and unlike so many writers lying about it, I will let people know what they're getting.

But I will not share that pen name with my self-written audience as AI haters will destroy my ratings on my other books.

I'm going to use it to write stories that are at the bottom of my pipeline and would probably never get written at the speed I write and publish myself. I want this project to be as much AI created as possible, and want to see if I can get the formatting, cover design, etc from AI.

After several attempts, I'm resorting to a process that requires me to iterate chapter to chapter. The AI loses frame of reference on which characters know what, the background of some characters, is wrong on its research usage at times, and loses POV.

The AI does a lot wrong but it can now generate 80% accurate/entertaining prose that I can clean up and embellish in one or two quick passes. I edit a chapter and then feed the book up to that point into the next chapter prompt. It is taking some time as I change up the plot and go back to rework chapters. Sometimes things don't occur to you until you read it.

I have not had 40 hours straight to work on it, but I can envision finishing a 300 page book in 3 weeks if I can focus. Even if I can't do most of the publishing aspects via AI, I think I can publish this book for $500-$1000.

One cool aspect of this is that I'm getting to read the book like one of those pick your own adventures, each chapter is new to me even though I know the major beats.

I will finish this first book and publish it, then I will add an AI day to my writing schedule, 3 on the trilogy, one on the rom com, and one on the next AI book. When I'm done writing the trilogy, we'll see how many books in total I will have to publish.

---

^--- Post written by ChatGPT?

---

Funny, nope and incredibly written by hand in my iPhone. I woke up and could not go back to sleep.

Recursive learning ai is hitting a wall, all have failed and demise into hallucination. Of course there will be improvements and then we get into a scenario where a super ai evolves and hits super hallucinations. That is the point when Terminator is chasing your ass to kill you.
Let's give the ai a weapon and get a defense contract is the problem.

ChatGPT wrote my mid-year review today.