Mind & Memory
Understand the person, protect context, reason honestly, and preserve a clean conversational mind across every model.
Progress that can be challenged.
Every node is a real program capability. It unlocks only when every acceptance criterion has merged evidence—and a failed regression can lock it again.
16%Capability readiness
7 of 45 fully unlocked
7 of 45 fully unlocked
70%
Scholar
Turn private books and approved sources into clear teaching, cross-source synthesis, and honest current awareness.
Analyst
Use approved data, deterministic execution, and typed evidence to answer analytical questions without inventing results.
Builder
Understand, change, test, and document approved local software projects with bounded permissions and inspectable evidence.
Creator
Turn conversation into excellent writing and artifacts through requirements, templates, rendering, validation, and review—not raw model output alone.
Personal Companion
Know the user without becoming intrusive, make work feel calm and coherent, and coordinate approved assistance as one loyal local companion.
Model Steward
Appreciate each open model, qualify its real strengths and limits, and route the smallest capable local intelligence for the work.
Guardian
Protect local ownership, permissions, provenance, truthfulness, evaluation integrity, and every boundary where information could leave the machine.
Open Platform
Make capable local AI installable, testable, collaborative, maintainable, and honest on the ordinary machines people already own.
52% development progress is not 52% product readiness.
Partial criteria receive half-credit in the secondary development metric. The primary readiness figure counts only capabilities whose complete gate is verified. Evidence was last reconciled on 2026-08-10.
The scoring covenant
No victory by declaration.
Private answers remain private. Public evidence points to reproducible suites, merged implementation and clear limitations.
Criteria, not code
A merged feature proves delivery, not mastery. Every listed behavior must pass.
Regressions count
A formerly dependable capability can return to training when a gate fails.
Credit stays
Contributor achievements remain attached to the people and evidence that earned them.
Complete audit
Inspect every node and checklist.
The repository contains the generated criterion-level Path to Mastery, evidence registry and contributor governance.