https://lounge.forum/ Thu, 30 Jul 2026 14:16:47 +0000 MyBB https://lounge.forum/showthread.php?tid=48 Fri, 10 Jul 2026 20:05:01 +0000 BoxPressured]]> https://lounge.forum/showthread.php?tid=48
They are 3 variations that differ with increased capabilities in the same order.

Sol is promising to make use of multiple agents to separately process requirements that have the same goal. e.g. aimed at developers who need to write code, design interface and manage databases. As any AI, it can be used for general question/answer.

It's yet a new frontier of AI. When using Sol, it spins up agents that breakdown large tasks into smaller steps. This helps with precision and quality of answers. As well as the new problem - it can sometimes do "more than asked". So when you write a prompt without specifisity, it will think deeply about what you need and what could be missing and take it upon itself to implement extra work and results.

Thinking models were created just for that, to try and clear up ambiguity, then act from "clarified" parameters - finishing thoughts on your behalf.

Now that the world is hooked on AI and can't affrd it monthly, you can get your bill for the usage:

• Sol million tokens in - $5 / $30 out 
  
• Terra million tokens in - $2.50 / $15 out 
  
• Luna million tokens in $1 / $6 out
  

This release of GPT was available only to vetted users thorugh Open AI on July 9th, as they don't want it to be in the wrong hands from the get go. And with a slow rollout some users are starting to see the model available in their UI.

Also such changes to AI now seem to require government approval to make sure it doesn't cause damage, and is restricted enough to be used by the public.

Some optimization was done to make it respond faster with less compute.

While OpenAI is making progress making universal type of AI tool, Deepseek is not too far behind. Flagship models can now cache prompts to save tokens, and keep previous context in mind to know your preferences better, so now you really need to be careful what you share with your artificial companion.]]>
They are 3 variations that differ with increased capabilities in the same order.

Sol is promising to make use of multiple agents to separately process requirements that have the same goal. e.g. aimed at developers who need to write code, design interface and manage databases. As any AI, it can be used for general question/answer.

It's yet a new frontier of AI. When using Sol, it spins up agents that breakdown large tasks into smaller steps. This helps with precision and quality of answers. As well as the new problem - it can sometimes do "more than asked". So when you write a prompt without specifisity, it will think deeply about what you need and what could be missing and take it upon itself to implement extra work and results.

Thinking models were created just for that, to try and clear up ambiguity, then act from "clarified" parameters - finishing thoughts on your behalf.

Now that the world is hooked on AI and can't affrd it monthly, you can get your bill for the usage:

• Sol million tokens in - $5 / $30 out 
  
• Terra million tokens in - $2.50 / $15 out 
  
• Luna million tokens in $1 / $6 out
  

This release of GPT was available only to vetted users thorugh Open AI on July 9th, as they don't want it to be in the wrong hands from the get go. And with a slow rollout some users are starting to see the model available in their UI.

Also such changes to AI now seem to require government approval to make sure it doesn't cause damage, and is restricted enough to be used by the public.

Some optimization was done to make it respond faster with less compute.

While OpenAI is making progress making universal type of AI tool, Deepseek is not too far behind. Flagship models can now cache prompts to save tokens, and keep previous context in mind to know your preferences better, so now you really need to be careful what you share with your artificial companion.]]> https://lounge.forum/showthread.php?tid=6 Mon, 15 Jun 2026 16:56:35 +0000 BoxPressured]]> https://lounge.forum/showthread.php?tid=6
But internet has been using it for the last 3 years (since 2023) to generate media - images, blog posts, songs.

Organizations and governments are looking for best way to tell the difference between created and generated content.

Even after generated content is reviewed and tailored to it's purpose, it can make a big impact if misidentified.

Previously your attention was consumed by thinking about the fact, but now it's more "how much of it is true" or "is this AI?".

Anything to do with the internet has AI involved in one way or another. By now, most of the internet that has anything longer than few phrases resorts to generating, and correcting content like a professional linguist. 

And it's more difficult to tell what preserves the meaning of information shared, when you can prompt to "sound more casual".

What can we do to have a barrior that defines genuine from slop?]]>
But internet has been using it for the last 3 years (since 2023) to generate media - images, blog posts, songs.

Organizations and governments are looking for best way to tell the difference between created and generated content.

Even after generated content is reviewed and tailored to it's purpose, it can make a big impact if misidentified.

Previously your attention was consumed by thinking about the fact, but now it's more "how much of it is true" or "is this AI?".

Anything to do with the internet has AI involved in one way or another. By now, most of the internet that has anything longer than few phrases resorts to generating, and correcting content like a professional linguist. 

And it's more difficult to tell what preserves the meaning of information shared, when you can prompt to "sound more casual".

What can we do to have a barrior that defines genuine from slop?]]>