code wiki / _hdl_build / nx_research_pipeline.nx
nx_research_pipeline.nx
buildroot/runtime/_hdl_build/nx_research_pipeline.nx
about
nx_research_pipeline.nx -- the governed research->task loop, fully wired (the deterministic next
move the layered backlog named). A finding only becomes a BUILD TASK if it survives BOTH gates:
GATE 1 (RESEARCHER): cross-source 3-vote adversarial verify -> CONFIRMED?
GATE 2 (CRITIC): scientific evaluation -> not REFUTED by a counter-theory, and actually
RED-TEAMED (not a blind spot)?
This is a LAYERED DEFENCE: the Critic is the DEEPER gate that catches what the Researcher's
source-agreement passed -- e.g. PSNR-as-quality is researcher-CONFIRMED (everyone uses it) yet
critic-REFUTED (Blau-Michaeli counter-theory), so it is correctly REJECTED. Dispositions:
TASK -> survived both gates + actionable -> placed at its layer for the backlog.
FLAG_REDTEAM -> confirmed but never red-teamed (a blind spot) -> Researcher hunts counter first.
REJECT -> researcher-unconfirmed or critic-refuted -> recorded, never built.
Then the surviving tasks feed nx_layered_backlog: the Conductor dispatches the deepest-layer
parallel batch. RACI: Researcher verifies, Critic red-teams, Conductor dispatches, Builder builds.
license_tier: ORIGINAL
dependencies 3 imports · 1 importers
imports: nx_researcher.nxnx_critic.nxnx_layered_backlog.nx
imported by: nx_research_pipeline_test.nx
structs
| none |
consts
| 20 | const RP_TASK: i64 = 1 |
| 21 | const RP_FLAG_REDTEAM: i64 = 2 |
| 22 | const RP_REJECT: i64 = 3 |
functions
| 25 | func rp_disposition(cv: i64, rv: i64, reproductions: i64, counters: i64, redteamed: i64, actionable: i64) -> i64 |
| 35 | func rp_run(n: i64, cv: *i64, rv: *i64, repro: *i64, counters: *i64, redteamed: *i64, actionable: *i64, layer: *i64, task_layer: *i64, counts: *i64) -> i64 |