Phase v0.1 — gate and recognition (Q2 2027)

On-device gate model (open-source, auditable).
Cloud recognition for three verticals: GeraEats (storefronts), GeraHome (appliances), GeraSure (vehicles).
Quote-card UI with explicit confirm.
Refusal policies enforced (face, minors, private env).

Open question: gate-model false-positive rate. If too many irrelevant frames reach cloud, privacy posture weakens.

Phase v0.5 — consent and commit (late 2027)

Full pipeline including GeraNexus consent + escrow + signed commit.
GeraMind vault integration for user preferences.
Audit-log dashboard for users.
Bystander-blur pipeline with public audit of false- negatives.

Open question: unit economics of the pipeline at scale. Cloud recognition per-frame cost needs to be under a few pence to support small-transaction verticals.

Phase v1.0 — breadth and enterprise (2028)

Five additional verticals: GeraRide, GeraRent, GeraMarket, GeraJobs, GeraLearn.
Regional visual vocabulary (10+ markets).
Enterprise SDK for third-party integrations.
v1.0 spec freeze with 2-year stability commitment.

Open question: whether enterprise demand justifies the regional-vocabulary expansion cost. Early conversations suggest yes; evidence pending.

What we will not do in v1

Face recognition. Ever.
Always-on background camera.
Commerce actions with minors as primary subject.

How to follow

Draft spec: geralens.com/spec. GitHub issues for debate. Email [email protected] if you work on ARKit, MediaPipe, or visual-commerce projects and want to collaborate.