Co-Author AI Autonomous writing agent

A disciplined AI co-author that watches you write, understands your intent, and contributes only when it truly adds value.

Comment

posted an update —

Project Update: Model Migration & Infrastructure Shift

Date: February 2, 2026 Status: Active Migration Complete

I have officially migrated the core engine of this project from the now-deprecated gemini-2.0-flash-exp to the latest gemini-2.5-flash-native-audio-preview-12-2025.

The Transition

With the shutdown of the 2.0 experimental endpoints in late 2025, I’ve moved the project over to the Gemini 2.5 Native Audio architecture. This was a necessary move to ensure the project remains functional and takes advantage of the most current flagship audio capabilities.

What’s New?

  • Native Audio Fidelity: The move to 2.5 brings significantly higher-quality audio synthesis and better emotional inflection in the model's voice.
  • Improved Multimodal Understanding: This model is specifically tuned for the Live API, offering better synchronization between audio inputs and generated responses.

Note on Latency

If you have used the previous version of this app, you might notice a slight change in response timing.

  • Previous Model: gemini-2.0-flash-exp was an experimental build optimized for raw, "bleeding-edge" speed.
  • Current Model: gemini-2.5-flash-native-audio prioritizes audio synthesis quality and complex dialog handling.

I am currently working on optimizing the system instructions and reducing the context overhead to bring the "Time to First Byte" back down to those ultra-low 2.0 levels.

Next Steps

I am monitoring the performance of the Gemini 3 lineup as it matures to see if a hybrid approach (using Gemini 3 for logic and 2.5 for audio) might further optimize the experience.

Thanks for your patience as I fine-tune the "soul" of this project with these new models!

Log in or sign up for Devpost to join the conversation.