The 12 Problems
A candidate taxonomy I am actively testing for an irreducible thirteenth problem.
Explore the taxonomy →Across people, systems, workflows, and AI, I keep the path from evidence to decision visible and governed.
AI can produce a next step without showing why it is justified—or whether it has authority to act.
I am developing ReasoningArc to make the evidence, judgment, uncertainty, and permission behind consequential decisions inspectable.
Personal research project; no software, certification, consulting, or other services are offered for sale.
A system may evaluate or recommend without permission to decide or execute. Before commitment, current authority must be checked; missing evidence or authority means pause, abstain, or escalate.
I intend this site for practitioners shaping AI-enabled processes, reviewers who need to inspect decisions, and researchers willing to challenge the architecture.
ReasoningArc is an evolving research framework and working architecture. It is not a certification, regulatory standard, or proven safety guarantee.
Each page separates established working requirements from hypotheses, observations, and questions that remain open.
A candidate taxonomy I am actively testing for an irreducible thirteenth problem.
Explore the taxonomy →How context, evidence, evaluation, authority, action, escalation, and records connect.
Follow the architecture →Recorded tabletop runs, limitations, and one illustrative end-to-end example.
Inspect the test record →A document-control inconsistency was identified in the ReasoningArc Decision Authority Standard v0.2 during the September 30 adversarial review. The header and change log correctly identified v0.2 / September 22, 2026, while the Document Control table retained v0.1 / August 10, 2026 metadata.
The controlled source has been corrected and verified. The historical finding remains in the Change Ledger as a resolved governance defect rather than being removed.
See the retained change record →