I admit that I've only searched for about 10-15 minutes, but I can't get a clear answer about what kind of AI content I can put up on Amazon. Is that the main distribution network being used or are you guys publishing to already acquired fans? Just curious as I'm trying to figure out what direction to go with my creative energies. I've got two series that I want to finish, but no time to write all the words of the books I've got outlined and planned.
To clarify, I see that Amazon takes "refined" with no specification but requires you to disclose "generated" which you then can't copyright. (Why is Amazon paying anybody for content that isn't copyrighted?)
> API access through platforms like OpenRouter or OpenAI's direct platform reveals GPT-5's true capabilities.
This is more or less true for every frontier model, though. I know you know this, VD, but consumers and light tech enthusiasts who aren't connected to how any of that works have zero idea or concept of what these LLMs offer and where they're most heavily used.
I noticed a while ago the better the functionality gets with all the models, the more disappointed some people are who never consider or understand what differences are in paid plans.
In my view, the biggest drawback to Claude Opus (other than price) would be the lack of poetic, atmospheric writing. I’ve been playing around with Claude Sonnet, and I asked it to write a 2,000-word scene three different times, each time asking it to mimic the style of an ornate prose stylist (Tanith Lee, Walter Pater, Bruno Schultz). The results were impressive. I suspect Claude Opus could overcome the problem of bland prose if one asked it to imitate the style of a particular writer.
Claude for creative and GPT for technical. One very serious aspect of AI that's not talked about is reverse engineering and it's a gamechanger. I can input a matrix chock full of magic numbers and get a very good guess of what transformation it's used for.
Recently did a 135k word novel with Claude 4.1 Opus. I did consistently hit the conversation limit, and the prompts for later chapters could've done with more reminders to read the worldbuilding docs I uploaded (eg eye colours changing), but on the whole I've been very pleased with it.
I made this switch in the 4o 04mini time of chatgpt and agree Claude adheres to instructions better.
Thats my primary requirement since "creative coding" is just replacing the last improvement with random changes and bad advice acted on without mention.
"Temperature settings matter significantly for GPT-5. Testing at lower temperatures (0.7 instead of 1.0) reduced the poetic tendencies and improved instruction adherence, though it also diminished some creative spark."
Wow, that's interesting. Lots of interesting implications there. I don't have access to Claude or Chat at the moment; is temperature something you can actually control for, or is that simply based on the activity of the servers at any given moment?
As far as tradeoffs between cost and creative quality go, based on what I've read here if I were to take the novel plunge with an AI assist I'd probably try to use both - Chat as the seed of an idea, then Claude to shape it into something more grounded. But if I had to choose just one, it would be Claude. If the story is worth telling, it's worth being told well.
I’ve been using Grok for outlines and basic drafts in lieu of ChatGPT, which I’ve rejected along with Gemini as too woke, but otherwise I use that strategy.
You can control it through an API like OpenRouter, which is really interesting to tool around with, I don't know about in. Claude directly as I don't have a subscription there.
That's fascinating, who would have thought a slight change in temperature could cause the "thought" processes in circuits to either become more rigid or more fluid?
I was wondering if that’s what it meant, kind of like warmer or cooler temperatures for photo adjustment. Then I wondered if there was a standard temperature setting range that people use. Thank you.
Okay, with a bit of research it makes perfect sense, but I'm a little disappointed that the way to change the type of response has nothing to do with the actual physical temperature of the system. Mainly because I know that keeping them cool is a huge issue, and the thought of overheated AI servers coming up with increasingly nonsensical answers as the temperature rises is highly entertaining. Plus it would be an opportunity to adjust prices on a sliding scale depending on how much coolant you use for their services.
What about using GPT-5 for lyrics writing compared to Claude 4 Sonnet?
I admit that I've only searched for about 10-15 minutes, but I can't get a clear answer about what kind of AI content I can put up on Amazon. Is that the main distribution network being used or are you guys publishing to already acquired fans? Just curious as I'm trying to figure out what direction to go with my creative energies. I've got two series that I want to finish, but no time to write all the words of the books I've got outlined and planned.
To clarify, I see that Amazon takes "refined" with no specification but requires you to disclose "generated" which you then can't copyright. (Why is Amazon paying anybody for content that isn't copyrighted?)
> API access through platforms like OpenRouter or OpenAI's direct platform reveals GPT-5's true capabilities.
This is more or less true for every frontier model, though. I know you know this, VD, but consumers and light tech enthusiasts who aren't connected to how any of that works have zero idea or concept of what these LLMs offer and where they're most heavily used.
I noticed a while ago the better the functionality gets with all the models, the more disappointed some people are who never consider or understand what differences are in paid plans.
In my view, the biggest drawback to Claude Opus (other than price) would be the lack of poetic, atmospheric writing. I’ve been playing around with Claude Sonnet, and I asked it to write a 2,000-word scene three different times, each time asking it to mimic the style of an ornate prose stylist (Tanith Lee, Walter Pater, Bruno Schultz). The results were impressive. I suspect Claude Opus could overcome the problem of bland prose if one asked it to imitate the style of a particular writer.
Yeah if you tell it a style it definitely helps in prompts.
Claude for creative and GPT for technical. One very serious aspect of AI that's not talked about is reverse engineering and it's a gamechanger. I can input a matrix chock full of magic numbers and get a very good guess of what transformation it's used for.
Recently did a 135k word novel with Claude 4.1 Opus. I did consistently hit the conversation limit, and the prompts for later chapters could've done with more reminders to read the worldbuilding docs I uploaded (eg eye colours changing), but on the whole I've been very pleased with it.
Opus burns through credits a lot faster than Sonnet. I prefer Sonnet for writing, Opus for research.
I'm an Anthropic man myself. Give me reliable text and I don't care if it costs a little more. Both systems are eminently affordable.
I made this switch in the 4o 04mini time of chatgpt and agree Claude adheres to instructions better.
Thats my primary requirement since "creative coding" is just replacing the last improvement with random changes and bad advice acted on without mention.
A tad off topic, would it be possible to enable Chat for AI Central?
We can take a poll tomorrow. If it would add value in the subscribers' eyes, why not?
Looks like a use both. I use hemmingway app and grammerly for editing.
I also use grammarly on editing
"Temperature settings matter significantly for GPT-5. Testing at lower temperatures (0.7 instead of 1.0) reduced the poetic tendencies and improved instruction adherence, though it also diminished some creative spark."
Wow, that's interesting. Lots of interesting implications there. I don't have access to Claude or Chat at the moment; is temperature something you can actually control for, or is that simply based on the activity of the servers at any given moment?
As far as tradeoffs between cost and creative quality go, based on what I've read here if I were to take the novel plunge with an AI assist I'd probably try to use both - Chat as the seed of an idea, then Claude to shape it into something more grounded. But if I had to choose just one, it would be Claude. If the story is worth telling, it's worth being told well.
I’ve been using Grok for outlines and basic drafts in lieu of ChatGPT, which I’ve rejected along with Gemini as too woke, but otherwise I use that strategy.
You can control it through an API like OpenRouter, which is really interesting to tool around with, I don't know about in. Claude directly as I don't have a subscription there.
That's fascinating, who would have thought a slight change in temperature could cause the "thought" processes in circuits to either become more rigid or more fluid?
Just to be clear, the temperature mentioned here is a label for a number you tell AI to use, not the temperature of the circuits running the AI.
I was wondering if that’s what it meant, kind of like warmer or cooler temperatures for photo adjustment. Then I wondered if there was a standard temperature setting range that people use. Thank you.
Okay, with a bit of research it makes perfect sense, but I'm a little disappointed that the way to change the type of response has nothing to do with the actual physical temperature of the system. Mainly because I know that keeping them cool is a huge issue, and the thought of overheated AI servers coming up with increasingly nonsensical answers as the temperature rises is highly entertaining. Plus it would be an opportunity to adjust prices on a sliding scale depending on how much coolant you use for their services.
Ah - very important clarification. Thank you. Now I feel like I need to do research.
On an only slightly related note, anybody here know how to retrieve data from a drive that isn't recognized? Some detail here: https://justsittingquietly.substack.com/p/bootless?r=1fcinz