Caliber tops 150 stars, auto-generates AI context files

Caliber is less a generator than a governance layer for AI instructions, and that is exactly where this category needs to go. Most teams do not need another tool that spits out `CLAUDE.md`; they need something that keeps those instructions honest after the repo changes.

• –The audit -> diff -> review -> backup -> undo loop is the right trust model for tooling that rewrites developer-facing config.
• –Local scoring and project-fingerprint generation make the privacy story much stronger than generic repo-upload tooling.
• –Multi-assistant support plus MCP discovery turns it into an onboarding layer, not just a file generator.
• –Session learning and auto-refresh hooks are the difference between a clever bootstrap script and a system that can actually stay current.
• –The space is getting crowded, so the moat will be drift detection and inference quality; 150+ stars is a healthy signal that the pain point is real.