Many voicebots these days need the speed of a fast LLM but the reasoning capabilities of a heavier LLM. To achieve this, modern products are shifting to a dual-model framework. An example is the recent GPT-Live release where a realtime speech-to-speech model has access to GPT5.5 in the background.

This is not a new design, but most of us simply use a heavier model in the background as a tool instead of trying out advanced techniques. For this issue of Computational Conversations I design a few of these techniques to use a fast foreground and slow background model.

I also reproduce a common speech-to-speech pattern of injecting knowledge from a larger LLM live midway (example from Sakana AI) in bot utterance without training a new model, which might be interesting if you like exhausting engineering options before going to model training.

Read more in the full post here.

Hlo,

Certain personal commitments have been keeping me from giving time to the newsletter. Until now.

If you recall, we touched upon how most web chatbots are expensive but poorly designed tech projects in the last issue. I came across a news coverage that shows just how costly these projects could be. The NYC city AI Chatbot costed $600,000 just to build and many agree, including the new Mayor, that it was “functionally unusable“. This is not counting the ongoing maintenance costs where, allegedly, most of the improvements and fixes would come from.

Hlo,

This is the first email from my small monthly newsletter with writings on Conversational AI. I will try to cut through the hype and talk about real world implications of Conversational Systems.

The first piece looks at web chatbots and argues that most of them don’t add new capabilities. They repackage existing website functions behind a conversational layer while disclaiming responsibility for errors. The result is often worse UX, unclear accountability, and new privacy risks.

But there is a meaningful role for chatbots as Assistive Technology. Read the full post here.