New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
In a multi-AI consult, even DeepSeek V4 brings unique findings to the table from my experience picking up issues even SOTA models didn't (confirmed by SOTA models). My multi-AI consults on average surface 50% to 200% more findings than standalone Claude Fable 5 or GPT-5.6 Sold can find!
I mainly use the multi-AI consults at planning or post-implementation verification phase so relatively they don't use as much tokens but ensure you start with an accurate plan to begin with = saving you tokens and time - preventing you from going down the wrong path.
Yes but reasoning is expentional the more models, 3 is intensive, 4 is very intensive, 8+ is stupid, or they aren't all reasoning with each-other and it's selective which is just as random as not having that many.
"Unique Findings" no was just that particular prompts which founds some of that particular billions of parameters at that time, all other models capable.
I assure you this is a waste of reasoning and LLM compute, it's beyond useless to have more than 4 models consult.
**4 models create 6 pairwise relationships; 8 create 28. **
Simply put, if it doesn't over-comptue that they aren't all reasoning with eachother anyways.
Well I don't use all the LLM models all the time, I said up to 8 LLM models supported. I usually only use 2-4 of them, Claude, GPT-5.6 Sol, GLM-5.3 are the main 3 right now.
I recall @layer7 has one of the best practice when responding to ticket
Too bad image that @sillycat post already expired
I think part of the “getting old” feeling is noticing other people behaving in ways you wouldn’t have considered acceptable yourself.
And I swear it’s not just me becoming a grumpy old man, staring at a wall of AI-generated text and muttering that the young whippersnappers should go back to actually debugging things (yet)
how do you feel about unboxing videos?
Hi,
the general practice with any kind of non technical what ever bla bla blob ( no matter if its coming from the internet / forum / blog or AI ) is that the longer the text is, the less brain activity will be done and more and more frequently whole lines of bla bla will just be skipped by my eyes and the brain will filter more and more the rest.
When i reached the buttom of this and did not understand it, i ask for clarification and more technical details :-)
EDIT:
But to contribute something more real here:
@SplitIce yes. The future is agents handling things. If they cant, the senior level agent handles ( Token will be more expensive ). And that goes up until some human will read it.... maybe