Product Digital Life Foundation
GHW’s product foundation goal: give a software product an evolution kernel that is auditable, approval-gated, rollback-capable, and continuously learning.
It does not let an agent directly modify a production system. Instead, it enforces a closed loop:
ProductIntent
-> KPI
-> ChangeProposal
-> Consent
-> Apply / Rollback
-> Audit
-> Existence Degree
Capabilities Already in Place
| Capability | Status | Entry point |
|---|---|---|
| ProductIntent loading | Implemented | product_intent.example.crm.json, --product-intent |
| KPI feeding into E | Implemented | KpiBridge, ghw_product_kpi_feed |
| Change proposals | Implemented | ProductEvolutionRuntime.proposeChange |
| CRM canonical proposal | Implemented | proposeCrmFollowupTightening |
| Approval / denial | Implemented | approve, deny |
| Apply / rollback state machine | Implemented | applyApproved, rollbackApplied |
| Audit log | Implemented | auditJson, /api/product/audit |
| Persistence & recovery | Implemented | product_runtime state block |
| Real product Adapter | Planned | Requires host system implementation |
Runtime Structure
ProductEvolutionRuntime holds:
product: the read-only ProductIntent.bridge: the KPI Bridge, which turns raw product metrics intoK.changes: the list of product change proposals.audit: the log of KPI refreshes, proposals, approvals, applies, and rollbacks.next_change_id/next_audit_id: monotonically increasing IDs across restarts.
How to Enable
./zig-out/bin/ui_server \
--state /tmp/ghw-product-state.json \
--ui-dir ui \
--product-intent product_intent.example.crm.json
Once enabled:
saveStatesaves the"product_runtime"block.changesandauditare restored on startup./api/product/*endpoints are available.
Boundaries
Currently, applyApproved means “GHW has entered the applied state and written the audit entry”; it does not directly modify the host product. Production integration requires a product Adapter that maps approved changes onto:
- Configuration centers
- Feature flags
- Workflow rules
- Experiment platforms
- Code generation / deployment pipelines
This is the focus of the next phase of production-grade product work.