Bound AI-assisted work
Stage work explicitly, preserve decision gates, constrain risky actions and verify outcomes instead of treating model output as completed execution.
Methods & Labs
20 published methods for bounded, reviewable AI work. Start from your problem, browse them by name or try one in a lab.
Composition, not method decoration
These methods are not trophies in a library. They are interoperable problem-solving components with separate responsibilities. A method earns its place by doing a bounded job: orienting a knowledge space, resolving identity or relationships, separating and grading claims, discovering existing capability, controlling execution, or verifying an outcome.
When a problem crosses several of those boundaries, the useful move is to compose the smallest compatible method stack and keep the hand-offs explicit. ORBIT can orient; SORR can resolve semantic identity and relationships; CARD & POINTER can represent and route; Dialogue Lifecycle can control when work advances; 5PP can discipline how bounded work is executed and audited. The value comes from those distinct responsibilities working together - not from naming more frameworks.
Relevance does not require ceremony. A task may need one method, several complementary methods, or none. If adding a method does not materially improve what is understood, decided, executed or verified, it is decoration rather than composition.
Find by problem
What is getting in your way?
Control and execution
AI work drifts, skips steps or acts before it should.
Bound work, decisions, instructions and risky actions.
See 3 methods →Reasoning and interpretationClaims, uncertainty or competing options are hard to judge.
Work with evidence, alternatives and independent baselines.
See 8 methods →Orientation and contextPeople or agents lose track of what things are and what to trust.
Resolve identity, provenance, relationships and relevant context.
See 4 methods →Discovery and designWe keep rebuilding what exists, or nobody owns the gap.
Find, reconcile and complete bounded capabilities before building more.
See 5 methods →All 20 methods
Plain names first, published method names underneath.
How work is bounded, staged and governed
Systems that control execution phases, dialogue transitions and instruction contracts.
5 Point Protocol
5PP5 Point Protocol with deterministic gates, compliance evidence and bounded execution discipline.
LifecycleConversation phases
Dialogue lifecycleEXPLORE, CLARIFY, DECIDE, EXECUTE, VERIFY and CLOSE as a separate when-control system.
Instruction contractsInstruction contracts
AICSVersioned instruction-contract specification with semantic atoms, execution gates, projections and conformance mechanisms.
How claims, discourse and alternatives are worked
Methods and representations for interpretation, inference, claim-strength control, possibility preservation and independent comparison.
Iterative reasoning loop
DIALECTICIterative heuristic-hermeneutic reasoning with bounded convergence and persistent knowledge units.
Semantic grammarSemantic grammar
RigVedanEBNF-RV grammar, temporal markers, semantic axes and typed knowledge units.
Interpretation methodInterpretation and teaching runs
Hermeneutic-didacticComparative profile of iterative interpretation and deterministic run pedagogy.
Claim reasoningClaim modes
DIAL-4Deduction, Inference, Abatement and Legitimation as separate claim modes.
Claim gradingClaim-strength grading
DIAL-4+Claim-strength grading across confidence, evidence quality and scope.
Possibility preservationKeeping options open
DIAL-4P PossibilityPossibility preservation and implementation governance without collapsing options too early.
Discourse decompositionDiscourse before intent
PISDPre-Intent Semantic Decomposition preserves discourse structure before reliable intent or task commitment exists.
Independent baselineIndependent first-pass baseline
IPBIndependent Predefinition Baseline preserves first-pass evidence before a developing definition can shape the comparison.
How subjects are identified, related and made understandable
Systems for provenance-aware orientation, semantic resolution, bounded representation, routing and context acquisition.
Orientation framework
ORBITTyped provenance-aware orientation nodes, routing planes, source-truth barriers and consumer projections.
Semantic resolutionSemantic resolution
SORRSemantic resolution for identity, representation, relationships, evidence, scope, time and authority while preserving uncertainty.
Orientation and routingOrientation cards and typed links
CARD & POINTERBounded semantic orientation plus typed reference and relationship routing with claim-scoped authority.
Context methodContext bootstrap
SRCBSubject Reconnaissance and Context Bootstrap builds bounded, source-aware context from distributed evidence.
How capability is found, reconciled, developed and used
Methods for avoiding false novelty, reconciling distributed capability representations, closing internal responsibility gaps and moving durable subjects into bounded use.
Reuse before building
Capability-Gap ReconnaissanceReuse-before-create discovery across distributed capability surfaces before new work is justified.
Capability reconciliationCapability reconciliation
CSRCapability Surface Reconciliation maps realization states, representation relationships, coverage and the smallest sufficient intervention.
Capability designResponsibility boundary design
CBECapability Boundary Expander closes internal responsibility gaps with KEEP, SPLIT, DELEGATE and REJECT while preserving peer boundaries.
Iterative capability designIterative boundary design
CBE-IXIterative Exhaustion follows a bounded expansion frontier to a semantic fixed point or an explicit stopping condition.
Discoverability and activationDiscoverability and activation
GDSAGoverned Discoverability, Surfacing and Activation separates durable existence, relevance, authorization, use and verification.
For a deeper view of how these capabilities translate into working software and operational evidence, continue to the proof packages, the source repositories or Adaptivearts.ai.
Practice
Six labs to start with
Interactive experiments using synthetic data. No account needed.
Blueprint reader lab
Rules, authority & enforcement
See why a written rule is not always an effective control. Includes CBE and CBE-IX exercises.
CBE / ClassificationComplete a capability boundary
Classify responsibilities, keep legitimate work and define handoffs to neighboring services.
CBE-IX / IterationKnow why iteration stopped
Follow an evolving design through gaps, handoffs, blocked decisions and identity drift.
5PP / Work protocolCheck a complete work record
Try 5 Point Protocol and distinguish a complete record from permission to proceed.
Dialogue LifecycleNavigate a dialogue
Explore when to clarify, decide, execute and verify during AI-assisted work.
Solution-space mapAssemble a method stack
Explore 11 recurring patterns and find a suitable combination of methods.
How the methods fit together: problems, roles and combinations
From methods to practical system capability
The value is not in isolated terminology. These systems combine into repeatable ways of designing, evaluating and operating AI-assisted work where context, uncertainty, authority and execution need to remain understandable.
Know what the system is reasoning about
Resolve identity, provenance, relationships and source authority across distributed repositories, documents, memory and runtime evidence.
Find and reuse what already exists
Discover, reconcile and activate existing capabilities before duplicating work or creating parallel implementations.
Keep uncertainty visible
Separate claim modes, preserve alternative interpretations and independent baselines, and resist premature intent or solution collapse.
Different problem shapes call for different capability combinations
The same methods can be combined differently depending on the problem, where it appears in the work lifecycle and the scale of the system. The inventory therefore doubles as a problem-solving map rather than only a list of named methods.
What is going wrong?
Ambiguity, weak evidence, context loss, semantic conflict, uncontrolled execution, duplicated capability or poor discoverability.
Where does it happen?
Observation, exploration, orientation, clarification, decision, execution, verification, learning or re-entry.
How large is the system?
A claim or conversation, a task or workflow, a project or repository, multi-agent work, or an ecosystem/portfolio surface.
The apparent task is not yet the real problem
Preserve discourse, tensions and alternatives before forcing a single intent, requirement or solution.
PISD + DIALECTIC + IPB + DIAL-4P
The system cannot tell what it is reasoning about
Reconcile identity, representation, relationships, provenance and source authority across distributed context.
ORBIT + SORR + CARD & POINTER + SRCB
Claims look stronger than the evidence warrants
Separate reasoning modes, grade claim strength and preserve independent evidence before definitions harden.
DIAL-4 + DIAL-4+ + IPB + SORR
AI output is being confused with authorized action
Stage work, preserve human and system gates, define instruction contracts and keep execution distinct from verified outcome.
Dialogue Lifecycle + 5PP + AICS + GDSA
A new build is proposed before existing capability is understood
Discover what already exists, resolve representations and select the smallest sufficient intervention.
Capability-Gap + CSR + SORR + CARD & POINTER
A capability grows by absorbing its neighbors
Identify genuine internal gaps, split mixed responsibilities, define handoffs and distinguish bounded exhaustion from a blocked or interrupted analysis.
CSR + CBE / CBE-IX + SORR + Dialogue Lifecycle + 5PP
Useful capability exists but stays operationally invisible
Move durable knowledge or capability through discovery, routing and relevance without confusing visibility with authorization.
GDSA + ORBIT + CARD & POINTER + SRCB
Many sources must become one bounded, defensible understanding
Acquire subject context, resolve conflicts, iterate interpretations and keep claim support visible.
SRCB + SORR + DIALECTIC + DIAL-4 + IPB
Projects, repositories and representations stop being navigable
Separate identities, orient across surfaces, find reusable capability and reconcile what exists where.
ORBIT + SORR + Capability-Gap + CSR + GDSA
Several plausible interpretations or paths remain open
Preserve possibilities, distinguish types of reasoning and make uncertainty part of the decision surface instead of hiding it.
DIAL-4P + DIAL-4 + DIAL-4+ + DIALECTIC + IPB
The catalog is organized by function rather than development stage. Problem classes above are solution patterns: they show how capabilities compose without implying that every problem always requires the full combination.
The problem-space map is the entry layer. Continue to the solution-space map to see how the current capabilities assemble into reusable solution patterns across lifecycle position and operating scale.