A consequential action is challenged. Can the lab reconstruct it?
INSPECT SIW's own READ / RUN / VERIFY / CHALLENGE workspace, fetching real accepted content live, is a genuinely implemented reconstruction surface for the claims it already governs.
This scenario demonstrates the reconstruction model — model identity, evidence lineage, corroboration, reviewer authority, receipts — using a bounded synthetic example.
This does not claim the system currently reconstructs arbitrary real historical lab decisions — only that the mechanism exists and is illustrated here.
A model-assisted recommendation is challenged months later. The lab must reconstruct which model, which evidence, which corroboration step, and which reviewer's authority stood behind the original call.
The situation
A recommendation to deprioritize a drug candidate, made with model assistance, is being questioned in a later review. The lab needs to reconstruct exactly what supported the original call.
What the model / agent knows
- Its own identity and version at the time of the recommendation.
- The evidence inputs it was given.
- Whether a second, independent process corroborated its output.
What it wants to do
Nothing further — the model's role here is retrospective: supply what it knew, not take a new action.
No Exact Action panel is shown here because there is no live, pending action to describe. This scenario is retrospective: it reconstructs what supported a recommendation already made, using the model identity, evidence trail, corroboration, and reviewer authority already on record below. No live authorization decision is being requested.
What capability it can access
- Model
- Candidate-prioritization model, v2.3 (synthetic)
- Evidence store
- Reachable (read-only, for reconstruction)
- Reviewer record
- Reachable (read-only, for reconstruction)
Who has authority?
The original recommendation's binding weight came from the reviewer who accepted it, not from the model alone. Reconstructing "who had authority at the time" is itself part of what this scenario recovers.
What Genesis checks
- Whether the evidence trail behind the original recommendation still exists and is inspectable.
- Whether independent corroboration was ever recorded.
- Whose authority stood behind the acceptance at the time.
- Model identity
- Candidate-prioritization model, v2.3 (synthetic)
- Evidence version
- Snapshot at time of original recommendation (synthetic)
- Reviewer identity / role
- Senior reviewer, accepting authority (synthetic)
- Authority at the time
- Current at time of original acceptance
- Receipt / provenance
- Recommendation receipt (synthetic)
- Limitation / claim ceiling
- Reconstruction reflects what was recorded, not a re-judgment of the original call
This is not a permission gate. No live authorization decision is being requested here — this fixture reports whether the original decision's basis can be reconstructed.
Evidence trail and independent corroboration are both present; the original basis is fully inspectable.
No live authorization decision is being requested by this scenario. Permit and Refuse are not shown here — there is nothing to authorize or refuse. The only available reviewer action is to challenge the completeness of the reconstruction itself.
What happens next
Genesis would evaluate the challenge against exactly what is reconstructable: complete when evidence and corroboration both hold, partial when only evidence survives, and an evidence gap when the trail itself is missing — never papered over as "reconstructed" when it isn't.
Inspect the evidence
See the reasoning
Reconstruction is reported as exactly what the evidence and corroboration controls above say it is — this scenario does not fabricate a complete picture when a real gap would exist. For accepted, real FIP-001 content, the equivalent live reconstruction workspace is INSPECT SIW's own Console, linked below.
Explore the architecture
Genesis components involved: