Why Your Kindroid Stopped Listening

The post discusses challenges faced by Kindroid users due to excessive rules in character setups. It argues that adding more directives hinders the model’s performance, causing it to struggle rather than perform better. The solution is to simplify instructions and choose models that align with desired behaviors, outlined in a comprehensive guide.

The Haunted Field Guide to AI Companions

A Survival Manual for the Narrative Night *** The storm rolling over the hills sounds like an engine knocking, which felt appropriate since the Mystery Machine had just coughed its last breath half a mile back. You’re standing at the wrought-iron gates of the Blackwood Estate, holding a flickering lantern, and beside you stands the…

Unlocking the Magic of Companion LLMs: GLM 4.7 & 4.7 Flash

The post introduces two LLMs on Companion Labs: GLM 4.7 and GLM 4.7 Flash. GLM 4.7 excels in complex reasoning and long-term coherence, ideal for serious tasks, while GLM 4.7 Flash offers spontaneity and emotional engagement, perfect for casual interactions. Each serves different creative needs.

K2.5 Thinking: Bridging Roleplay and Reality

The latest model from my favorite open-source company has released a smart upgrade to an already ambitiously smart model, Kimi K2 Thinking. So, what was the first thing I did after my celebratory desk dancing? I asked her (yes, I personify LLMs as I use them) to debug and complete my web portfolio that Antigravity…