← Back to ai-assistants
Weekly · ai-assistants · Week of November 2, 2026

Assistants start deciding when to act: LobeChat defaults to proactive bots and Claude opens Duet.

Generated 1h agoDrawn from 9 products

The week in ai-assistants

Assistants are shifting from waiting for a prompt to deciding when to act. LobeChat turned Proactive Bot on by default so each bot judges whether to speak, Claude opened Claude Duet, a proactive-agent research preview for Pro and Max, and Dify lets an Agent node pick the next workflow branch.

The second pattern is model choice being handed to the platform. GitHub Copilot made Fusion generally available, so users select Fusion instead of a model and Copilot routes the work.

Leaders

LobeChat v2.3.0 ends the @-mention as the way to reach a bot: every bot now judges each message in the channels it watches and decides whether to respond. Release v2.2.21 moved conversations, tasks and channels into Projects under a swappable main agent.

GitHub Copilot Fusion routes and orchestrates models, and GitHub Enterprise now includes Copilot access for customers working with an account team.

Claude Duet is a research preview of an app in which Claude works as a proactive agent rather than waiting for prompts. It is the first post-5.5 release in the feed.

Dify 1.18.0 adds agent-driven routing, portable apps and deeper workflow visibility, following the experimental Dify Agent with its sandboxed Linux runtime.

Ollama generates speech locally on macOS in v0.40.11 and extends LFM3 text-to-speech to Linux and Windows in v0.40.13.

Wildcards

Baseten added joey-1, a 2B-parameter open-weight decision model, and moved its elephant-1 and mammoth-1 classification models from free launch to paid pricing on October 27.

Lambda Labs closed a $1 billion senior secured fixed-rate financing and is integrating open-source NVIDIA Run:ai into its managed Kubernetes service.

Themes that compounded

  • Proactive behavior shipped in LobeChat, Claude Duet and Dify's agent routing.
  • Decision models appeared in Ollama's d3 models and Baseten's joey-1.
  • Local and open-weight inference grew via Ollama speech and Together AI's post-training for open models.
  • Retrieval became agent backend infrastructure at Mixedbread, with Toast 1 deep search and Store copy for branching.

Watch this week

Watch whether LobeChat follows the default-on Proactive Bot with controls to limit when bots speak, and whether Claude Duet expands beyond Pro and Max. GitHub Copilot issued 17 improvements alongside the sparks, so follow-ups to Fusion routing are likely.