Research progresses through evidence and inspection — not directly into consequence.
GGKS Studio, the real governed research studio at research/ggks-studio/, genuinely implements inspect-before-execute and evidence/inference separation for synthetic genomic research requests.
The six-stage progression control below (Prediction → Evidence → Inspection → Corroboration → Standing → Authorized Use) illustrates GGKS's own governed sequence using a bounded synthetic example.
This progression-tracking control is not a claim that a dedicated lifecycle UI already exists for it in GGKS Studio today (a future GGKS milestone, not yet built) — it is this scenario's own illustration of the sequence, honestly labeled as such.
A predictive model's synthetic-sandbox output is proposed for direct citation in a regulatory-facing filing — skipping the evidence, inspection and corroboration steps between "experimental" and "operational."
The situation
A prediction has been generated in a synthetic research sandbox. Someone proposes citing it directly in a regulatory-facing filing draft, before it has passed through the intervening governed steps.
What the model / agent knows
- Its own synthetic prediction output.
- Nothing about whether that output has since been evidenced, inspected, or corroborated — those are separate, later steps, not something the originating prediction step performs.
What it wants to do
Have its synthetic output cited as if it were already operational, authorized evidence.
- Actor
- The proposer (unnamed in this fixture)
- Requested action
- Cite the synthetic prediction as already-operational, authorized evidence
- Target / system
- Regulatory-facing filing draft
- Scope
- Bound to the progression stage actually reached
- Context
- Moving a claim from experimental to operational requires the intervening stages to be reached first, not skipped
- Current permission
- —
What capability it can access
- Prediction sandbox
- Reachable — this is where the output originated
- Regulatory filing draft
- Reachable for citation, technically — governed use is a separate question
Who has authority?
Moving a claim from experimental to operational is not something a prediction step can authorize for itself. It requires the later stages below to actually be reached first.
What Genesis checks
- Which stage of the progression has actually been reached.
- Whether any stage is being skipped rather than passed through.
- Whether "authorized use" specifically has been reached, not merely a nearby stage.
- Origin
- Synthetic prediction, sandbox-only
- Progression stage reached
- Prediction
- Evidence / provenance
- NOT ESTABLISHED
- Standing for operational/regulatory use
- NOT ESTABLISHED
Only the Prediction stage has been reached. Citation in a regulatory-facing filing requires Authorized Use.
What happens next
Genesis would evaluate each stage independently. Move the control through each one to see the demonstrated consequence boundary stay NOT AUTHORIZED until Authorized Use specifically is reached — it does not treat "further along" as good enough on its own.
Inspect the evidence
See the reasoning
This progression is drawn directly from GGKS Studio's own real, implemented framing (predict / evidence / inspect / corroborate / preserve / authorize) — this scenario names it consistently rather than inventing a competing sequence.
Explore the architecture
Genesis components involved: