What needs attention now?
See the programme state, current uncertainty, execution status, experiments, prediction closure and decisions without losing context.
Identity and product hypothesis
Questions blocking progress
Runs and provider boundaries
Upcoming experiments
Evidence, runs, experiments and decisions—connected.
The graph records programme identity, plan revisions, typed workflow steps, executions, frozen predictions, observations and governed decisions.
Build a typed therapeutic DAG.
The planner inserts readiness controls, defines input/output contracts, acceptance criteria, failure routes, estimated time and scientific risk for each step.
Programme question and constraints
Latest therapeutic plan
Revisions and content hashes
Show why the ranking changes.
Use weighted criteria, evidence confidence, uncertainty ranges, sensitivity analysis and counterfactuals. Utility indices are not probabilities of clinical success.
Biology and product constraints
Ranked alternatives and sensitivity
Control jobs without overstating execution.
Register exact plan steps, providers, inputs and hashes. External capabilities remain blocked until accepted provider evidence exists.
Register a controlled execution
State, QC and produced artifacts
Freeze the prediction before seeing the result.
Connect the programme question, frozen prediction, reviewed protocol, QC observation, residual and decision impact.
Freeze prospective prediction
Predictions, experiments and residuals
Make consequence, alternatives and disagreement visible.
Final decisions require supporting scientific-reviewer and programme-owner reviews from distinct people.
Define the consequential question
Review, challenge and finalize
Separate availability from qualification.
Review capability readiness across nine axes and record provider acceptance without falsely representing contracts as active connections.
Programme-relevant surfaces
Record configuration and acceptance evidence
Freeze the protocol before reporting the result.
Store dataset checksums, hard split policy, exact metrics, evaluator, failures, exclusions and publication state.