Public Roadmap
This roadmap tracks the current public direction of DBaD as a governance protocol for trust over time.
It focuses on what is being stabilized, clarified, and made more truthful in the public system. The active staged review baseline is Round62 / DBAD-PUB-1067, held for continued local hardening before the next peer send, with active tool-boundary direct-run guard coverage.
- Public draft baseline
- Trace realism
- Runtime truthfulness
- Known limits remain explicit
Now
- Stabilize the public baseline across the site, explainer, and evaluator.
- Route review traffic through the Round62 current-state page, fixture suite, API docs, and peer-review entry surface before older peer history.
- Keep the live runtime truthful across evaluate, trace storage, validation, verification, and transition.
- Maintain public clarity around what DBaD solves and what remains a boundary condition.
- Collect real-world feedback, confusion points, and misuse patterns.
Next
- Continue the evaluator v2 backend payload rollout.
- Expose richer trace-native outputs in the public evaluator.
- Reduce the remaining front-end preview logic.
- Add more worked example traces or scenario walkthroughs where they improve understanding.
After
- Translate validated refinements into tighter runtime outputs where justified.
- Improve trace visibility for reset, verifier reuse, and transition boundaries.
- Refine teaching and reference surfaces only where real confusion justifies it.
Later
- Deeper runtime-enforcement realism.
- Internal research corpus and adversarial benchmark growth.
- Possible implementation-grade pilot surfaces once justified by the work.
Completed in the Current Public Slice
- Proof-backed examples linked to stored canonical traces.
- Deterministic trace validation.
- Action guidance on trace detail pages.
- Multi-model peer review findings.
- Try to Break DBaD challenge page.
- Logic-review reporting flow.
- First-pass DecencyMeter advisory scoring demo using one runtime-audited trace.
Implemented v2.2 Slices
outcome_statusis now live on stored traces.outcome_historyis append-only and records manual post-decision updates.- Outcome updates increment
trace_versionand refreshupdated_at_utc. escalation_closureandescalation_closure_historyare now live on traces with escalation context.declared_blind_spotsis live as a structured scope-limit field.expected_outcomeis live as a pre-committed expectation field.state_transition_evidenceand optionalevidence_hashare live on transition records.completeness_attestationis live as a trace-level coverage claim.- Trust-positive continuation now requires closure when escalation context applies.
- These runtime slices remain observational and governance-focused only. They do not certify truth, completeness, correctness, or safety.
For one public runtime-audited example of the full stack working together, open the DBaD v2.2 Runtime Demo.
For the architectural separation between trace protocol and downstream scoring, open the DBaD → DecencyMeter Bridge.
The first downstream scoring surface is now live at /decencymeter/demo. It uses audited trace trc_20260428181140_42396240, shows explicit subjective weights, and remains advisory rather than DBaD validation.
Future Research Hardening
The field stack is implemented. The remaining work is hardening, canonicalization, and larger-scope governance research.
- Evidence: stronger evidence canonicalization and more explicit hashing-input rules.
- Enforcement: selective evidence enforcement by transition type where justified.
- Completeness: cross-trace completeness or publication-window attestation, if later justified.
- Aggregation: pattern and integrity insights across outcomes, blind spots, closures, and evidence-linked transitions.
Roadmap Shape
This is not a feature-checklist roadmap. It is a roadmap for strengthening enforceable trust, trace visibility, and evaluator truthfulness over time.
Continue with the current-state baseline, public update notes, examples, the peer-review findings, Try to Break DBaD, review what DBaD solves, inspect the trust flow diagram, read the white paper v3, or use the research demo.