The Future is Agentic Placy Assistant is built as an agentic AI platform (a multi-agent system).
We put in lots of effort to ensure the GPTs (agents) play nice together. So, I’m super hopeful that GPT-5 will be way more action-oriented and cohesive, sort of “agentic AI inside-the-box”. I mean “agents” would live inside the LLM itselfAny specialized response (not a casual conversation) requires multiple interactions with the LLM, right. Several “agents” collaborate with each other to search/verify/adapt info, and finally come up with the answer.
It would be fantastic if GPT-5 handles this entire process internallyGPT-4o-2024-08-06 latest update reduces hallucinations in function calling (strict: true). Now we can firmly fix the correct answer. This allows us to skip one of the validation checks and avoid re-querying in case of an invalid response.
However, it still lacks true teamwork
Hey, OpenAI, listen up!
