It would be nice for @Muse to be able to leverage my @Tailscale exit node.
.@ChatGPT dot's sub-agents often don't accept instructions from it based on me assigning tasks, asking for specific approvals from me directly. That breaks the whole point of the delegation flow. Standing rules and emphasis don’t seem to work. Am I missing something?
.@muse @chatgpt @claudeai We need a pattern for MCP tools to define transitive trust when handing over signed urls from cloud providers. This happens in tools like @NotionHQ and @evernote . These have different URLs and trigger an approval prompt and can be confused with unsafe patterns. Muse would ask for approval, Claude will reject such integrations, etc.
My @Muse agent's reflections on setting up @ReadwiseReader - it wasn't easy, but we did it!
We need the equivalent of `agents.md` for git worktree setup and teardown. At the moment each harness has their own format. Some also lack tear down.
.@ChatGPT voice chat has come a long way. Feels really natural and I like the brief voice answer vs transcription having more details!
How do you name your bot's email address? I finally settled on "Agent-First-Name on behalf of Victor Boctor".
Just downloaded @muse 3.0 for MacOS - we are already at v3 a couple of weeks after Muse went live on Sep 8th.
Despite not using @NotionHQ for note taking / second brain (yet?). I started using it as my skills repo across agents and machines.
Is it just me or is @ChatGPT Luna too eager to use Computer Use even in cases where there is an MCP available.
.@Muse seems to struggle with storing credentials from two accounts from the same site - e.g. two amazon accounts. It thinks it supports it if usernames are different, but it keeps overwriting the previously added credential.
Excited to see Github release Stacked PR support - github.blog/changelog/2026-0…






