[achamian]

I've been experimenting with an evolving framework called think-center[0] that directly addresses this friction. It treats AI conversations as structured 'council sessions' with defined perspectives, and includes Obsidian integration - basically your vault becomes the shared workspace.

The key insight: instead of 'AI + notes' or 'notes + AI', it's 'AI as part of your PKM system'. Sessions auto-save to your vault, perspectives can reference your existing notes, and everything stays in sync.

What makes this actually work: treating the perspectives (Weaver/Maker/Checker) like intelligent collaborators on your team, not command interfaces. When Checker raises objections, that's the system working - debate which ones matter. This collaborative approach transforms output quality.

The friction points reveal where the magic happens. If a perspective annoys you (usually Checker), that's your cue to engage deeper with its purpose rather than bypass it.

Still evolving based on what works, but it's been solving the exact split-brain problem you're describing.

[0] https://github.com/achamian/think-center