Short-term roadmap
Near-term execution hypothesis / requires validationThe near-term priority is not to announce autonomy. It is to establish shared interfaces, evidence practices, and safe experiments that make deeper coordination possible.
Near-term workstreams
Each stream should produce something inspectable: a contract, a test result, a documented workflow, or a decision that narrows the next experiment.
Inventory project boundaries, service contracts, route surfaces, data ownership, and reusable capabilities that can safely be shared.
Define how research notes, model evaluations, operational lessons, and project documentation are captured, reviewed, and reused.
Test provider abstraction, model selection, credentials, usage accounting, failure handling, and policy controls before connecting sensitive workflows.
Use narrow internal experiments to test whether one project can provide useful capability to another without hidden coupling or unsupported claims.
Study whether Bharaka should become a venture or support layer, including mandate, governance, risk, funding sources, and conditions for stopping support.
These workstreams describe an order of investigation, not committed delivery dates. A workstream is complete only when its evidence and decision boundary are documented.
Experiment register
The portfolio can learn from itself, but only when experiments are isolated, consented, measurable, and reversible.
Can a bounded set of reviewed documents improve a model-assisted learning task without weakening source traceability?
Use non-sensitive documents, show citations, log model/provider use, and keep publication decisions human-reviewed.
Does a reusable framework or service reduce project effort without reducing project ownership or reliability?
Sandbox first, project-specific adapter, no production dependency until reliability and rollback are demonstrated.
Can recurring operational patterns be documented and reused without exposing private customer or financial data?
Share abstracted patterns only; no raw customer data, secrets, credentials, or unreviewed model output.
What would a responsible support or venture layer need to remain solvent without distorting project decisions?
Scenario modelling only until governance, capital source, risk appetite, and accountability are formally defined.
Decision gates