unit_id	axis	current_state	next_action	acceptance_gate	evidence_objects	unlocks
V95-DATASET-IMPORT	Dataset import	EXECUTABLE_SPEC_READY	Register candidate DMD single-cell, single-nucleus and spatial datasets with donor, disease, state, perturbation and checksum fields before browsing.	At least two disease-relevant datasets have stable accessions, donor-level metadata, raw/processed SHA-256, license and prohibited-use fields.	api/v1.1/dataset_registry.json;api/v1.1/competitive_upgrade_blueprint.json	Real Cell Browser v1 import queue; no prediction claim.
V95-CELL-BROWSER	Real Cell Browser	PLANNED_REQUIRES_DATA_IMPORT	Replace concept-only context cards with donor-filtered cell-state panels, expression lookup, pseudobulk differential tables and exportable selections.	Browser selections resolve to Dataset object IDs and never summarize disease effect without donor-aware aggregation.	/resource/cell-browser/;api/v1.1/dataset_registry.json	Inspectable cell-state substrate for future perturbation interpretation.
V95-MODELRUN	ModelRun adapter	EXECUTABLE_SPEC_READY	Run zero, train-mean, nearest-neighbour and ridge baselines under a shared immutable ModelRun envelope before any new model is promoted.	Each run records model id, commit, container digest, data release, split id, seed, metrics, logs and failure reason.	api/v1.1/model_registry.json;api/v1.1/benchmark_suite.json	Comparable baseline ledger; calibrated DMD model count remains zero.
V95-PERTURBATION-NETWORK	Perturbation delta network	CONTRACT_READY_DELTA_LOCKED	Bind candidate perturbations to pathway modules, cell-state transitions, functional readouts and the missing evidence edge that would unlock each delta.	Every edge declares observed, inferred, planned or locked state; layout seed is fixed and node position is not interpreted as biology.	api/v1.1/v95_execution_network.json;api/v1.1/perturbation_cascade_network.json	Network-shaped experiment design, not a pathway-delta prediction.
V95-PROSPECTIVE-BENCHMARK	Prospective benchmark	SPEC_READY_NO_HIDDEN_OUTCOMES	Freeze Study Card, Prediction Card and Outcome slots before laboratory outcomes are visible.	Timestamped prediction objects exist before outcome access, with null, toxic, failed-QC and abstained outcomes preserved.	api/v1.1/study_card_manifest.json;api/v1.1/prediction_registry.json	Future hidden-test discipline; no current P1/P2 unlock.
V95-OUTCOME-CHAIN	Outcome return path	OBJECT_SCHEMA_READY_NO_MEASURED_OUTCOME	Append measured outcomes to the same object chain that generated the study and prediction records.	Study -> Prediction -> Experiment -> Outcome IDs remain linked without regenerating downstream IDs from URL parameters.	/resource/workbench;api/v1.1/prediction_registry.json	Prospective evidence accumulation once real outcomes arrive.