EXPLORATORY RESEARCH LAYER

Flash is the exploratory research layer of GenerativeDriveOS. It accelerates external discovery and synthesis while keeping outputs explicitly provisional and separated from canonical truth.

Flash operates beyond the internal canon: web sources, external documents, external signals, and emerging context. Its outputs are not treated as truth by default — they are hypotheses, leads, and mapped evidence for review.

Flash surfaces novelty, drift, and alternative explanations against the organisation’s canon. It produces research packets: sources → notes → candidate claims → confidence labels, enabling fast exploration without contaminating canon.

Flash has no authority to modify or override internal canon. Outputs carry explicit epistemic labels (external / unverified / speculative) and decay via TTL.
Flash can highlight inconsistencies or external signals. It cannot determine truth or assign fault.

Flash runs as a controlled research console: query → web synthesis → labelled packets → optional promotion via Airlock. Promotion requires explicit human action and policy/evidence checks before anything can enter canon.

High-speed exploration, explicit uncertainty, fail-closed boundaries. No stable IDs for Flash artefacts, expiry by design, and strict separation from DKG — ensuring discovery remains safe, auditable, and non-authoritative.

Flash can be wrong by design. That is what makes it safe.