# AX Patterns Specification # Machine-readable trigger conditions for all 23 Context Grammar AX Patterns # Source: intentfirst.ai — context-grammar/ax-patterns/index.html # Version: 1.0 | Date: 2026-05-05 spec: name: AX Patterns Specification version: "1.0" date: "2026-05-05" description: "Trigger conditions and metadata for the 23 reusable AX Patterns in Context Grammar" license: "CC BY 4.0" url: "https://intentfirst.ai/context-grammar/ax-patterns" token_spec_ref: "./context-tokens-spec.yaml" design_rules_ref: "./design-rules.yaml" note: > Token values use enum IDs from context-tokens-spec.yaml. Brain conditions (L1/L2/L3), AI confidence, and system state are non-token inputs documented in trigger_conditions.extended. # ------------------------------------------------------------------- # Token value reference (for trigger condition validation) # ------------------------------------------------------------------- token_value_index: physical_state: [stationary_both_hands, stationary_one_hand, walking, transit_passive, driving, exercising, lying_down] cognitive_load: [low, moderate, high, overloaded] social_exposure: [private, trusted_partner, family_with_children, social_acquaintances, public] priority_weight: [critical, high, standard, low] form_factor: [phone_handheld, phone_folded, phone_unfolded, tablet, tv_display, fridge_display, car_display, watch, desktop_monitor] feasibility: [fully_feasible, partially_feasible, infeasible] autonomy_dial: [suggest, confirm, notify, auto] disclosure_dial: [none, minimal, moderate, full] # ------------------------------------------------------------------- # Pattern Categories # ------------------------------------------------------------------- categories: D: name: Delegation description: "Trust transfer from human to AI — the human hands authority to the AI" direction: human_to_ai count: 6 E: name: Escalation description: "Control return from AI to human — the AI hands the decision back" direction: ai_to_human count: 5 A: name: Adaptation description: "UI and content transform in response to context changes" direction: context_driven count: 8 X: name: Enterprise description: "Organizational patterns for multi-person, compliance-driven, or high-stakes contexts" direction: organizational count: 4 # ------------------------------------------------------------------- # D — Delegation (6 patterns) # ------------------------------------------------------------------- patterns: - id: D1 name: Approval Gate category: D essence: "A checkpoint where the user approves, edits, or rejects an AI proposal before it executes." trigger_tokens: autonomy_dial: confirm priority_weight: [high, critical] trigger_extended: description: "Fires when Autonomy is in Confirm mode and the proposed action has High or Critical weight." note: "D1 is the baseline delegation pattern — before Progressive Trust (D2) has elevated the dial." outputs: ui_pattern: approval_card primary_actions: [approve, edit, reject] edit_path: X2 # Inline Edit is the edit sub-pattern example: "The fridge proposes a grocery order — Hana taps Approve, Edit, or Skip before anything is sent." - id: D2 name: Progressive Trust category: D essence: "AI earns the right to act with less friction as successful outcomes accumulate over time." trigger_tokens: autonomy_dial: [confirm, notify] # Current dial before upgrade trigger_extended: brain_condition: layer: L2 signal: consecutive_successful_approvals threshold: "configurable — default 5" description: "Fires when Brain L2 records sufficient consecutive approvals without user correction." note: "Proposes dial upgrade to user. Does not auto-upgrade — user must confirm the first time." outputs: ui_pattern: upgrade_proposal proposed_transition: "confirm → notify, or notify → auto" example: "Weeks 1–3 Hana approves every grocery sub. By Week 8 the system moves to Notify — she sees the swap after, not before." - id: D3 name: Proactive Nudge category: D essence: "AI surfaces a non-urgent suggestion precisely when Cognitive Load is low enough to act on it." trigger_tokens: cognitive_load: low priority_weight: standard feasibility: fully_feasible trigger_extended: description: "All three conditions must be true simultaneously. Low load ensures the user has bandwidth. Standard priority ensures no interruption urgency." note: "If cognitive_load rises to moderate or high, the nudge is deferred — stored as candidate, not surfaced." outputs: ui_pattern: ambient_suggestion_card dismissable: true defer_on_load_rise: true example: "Saturday morning, nobody's rushing — the app surfaces 'Kai's soccer cleats need replacing, end of season sale now.'" - id: D4 name: Omakase Mode category: D essence: "Full delegation to AI in a domain where trust, disclosure, and Brain confidence are all at maximum." trigger_tokens: autonomy_dial: auto disclosure_dial: full trigger_extended: brain_condition: layer: L2 signal: domain_trust_level threshold: high description: "Requires all three axes at maximum: Autonomy=Auto, Disclosure=Full, Brain L2 trust high in this domain." note: "Per-domain — a user may be in Omakase Mode for grocery staples but Suggest for travel." outputs: ui_pattern: silent_execution post_execution: activity_log_entry user_facing: none example: "Hana says nothing — the regular grocery order runs itself, exactly as she would have placed it." - id: D5 name: Substitution Modes category: D essence: "Four levels of substitution freedom — Exact, Flexible, Exploring, Surprise — per item and per risk level." trigger_tokens: feasibility: [partially_feasible, infeasible] trigger_extended: brain_condition: layer: L1 signal: per_item_substitution_setting values: [exact, flexible, exploring, surprise] description: "Fires when requested item cannot be fulfilled as-is. Substitution mode is stored per-item in Brain L1 (stable preference)." outputs: substitution_modes: exact: description: "Same item, different source" autonomy_requirement: suggest flexible: description: "Close equivalent meeting most criteria" autonomy_requirement: confirm exploring: description: "AI-expanded alternatives user might not have considered" autonomy_requirement: confirm surprise: description: "AI autonomously picks best available" autonomy_requirement: notify example: "Koji's peanut butter is Exact (allergy risk). The yogurt brand is Flexible. A new snack is Surprise." - id: D6 name: Dynamic Friction category: D essence: "High-cost or high-risk actions keep a confirmation step even when overall trust is mature — a ceiling that Progressive Trust cannot exceed." trigger_tokens: autonomy_dial: [notify, auto] # Trust is already elevated... priority_weight: [high, critical] # ...but action risk is high trigger_extended: risk_condition: dimensions: [financial_cost, irreversibility, safety_impact] threshold: configurable description: "Fires when cost × irreversibility product exceeds threshold, regardless of Autonomy Dial level." description: "Overrides elevated Autonomy Dial for specific action types. Trust and risk are independent axes." outputs: ui_pattern: forced_approval_gate overrides: [D4] note: "Always generates a D1 Approval Gate for the matched action, even at Auto autonomy." example: "The fridge reorders yogurt automatically. But a ¥15,000 appliance always asks first — trust and risk are independent axes." # ------------------------------------------------------------------- # E — Escalation (5 patterns) # ------------------------------------------------------------------- - id: E1 name: Confidence Signal category: E essence: "AI displays its uncertainty in graduated levels — not hiding doubt, but making it legible." trigger_tokens: social_exposure: [social_acquaintances, public] # Higher exposure = more critical to be transparent trigger_extended: system_condition: ai_confidence_threshold: "≤ 0.80" description: "Fires when AI model confidence falls below 80%. Social exposure context determines how prominently the signal is displayed." description: "Not purely token-triggered — requires AI confidence signal from the model layer." outputs: confidence_levels: high: "≥ 0.80 — standard display, no confidence indicator" medium: "0.40–0.79 — confidence percentage visible, alternatives offered" low: "< 0.40 — escalates to E4 Ambiguity Escalation" example: "The travel app shows '71% match — alternatives available' for the hotel recommendation, not a confident single choice." - id: E2 name: Limitation Disclosure category: E essence: "AI honestly names what it cannot do and redirects to the appropriate resource." trigger_tokens: priority_weight: [high, critical] trigger_extended: system_condition: request_type: outside_ai_capability description: "Fires when request type is outside AI's defined capability boundary AND priority is high." description: "Priority condition ensures routine out-of-scope requests don't generate noise." outputs: ui_pattern: limitation_card required_elements: - explicit_capability_boundary_statement - redirect_to_appropriate_resource tone: "honest, not apologetic" example: "Mia reports feeling unwell mid-trip. The app says 'I can find nearby clinics, but I cannot make medical decisions' and shows a map." - id: E3 name: Rollback category: E essence: "AI reverses a completed action and offers alternatives when external conditions change after execution." trigger_tokens: feasibility: infeasible # Changed to infeasible AFTER execution trigger_extended: state_condition: execution_state: completed feasibility_change: post_execution description: "Fires when feasibility of an already-executed action becomes infeasible due to external change." outputs: ui_pattern: rollback_card required_elements: - visual_strikethrough_of_cancelled_item - pre_computed_alternative - cost_delta_if_relevant example: "The booked 14:00 flight is cancelled by the airline. The app strikes it through, shows a 16:30 alternative at no extra cost." - id: E4 name: Ambiguity Escalation category: E essence: "AI surfaces gray-zone decisions with a warning and explicitly defers the final call to the human." trigger_tokens: {} # No token trigger — driven by AI confidence only trigger_extended: system_condition: ai_confidence_range: "0.40–0.79" description: "Fires when AI confidence falls in the gray zone — too uncertain to commit, too relevant to discard." outputs: ui_pattern: ambiguity_card required_elements: - explicit_doubt_statement - two_or_more_options - deferred_decision_prompt note: "Below 0.40 confidence, escalation may be unnecessary — silence may be better than a low-confidence guess." example: "The planner isn't sure if Kai would enjoy a particular museum. It flags the doubt, shows two options, and says 'your call.'" - id: E5 name: Trust Breach Recovery category: E essence: "After an AI misjudgment, the Autonomy Dial is demoted and trust is rebuilt from a lower stage." trigger_tokens: autonomy_dial: [notify, auto] # Elevated dial that triggered the misjudgment trigger_extended: user_action_condition: action: [reject, undo] context: "AI-initiated action or proposal" description: "Fires when user rejects or undoes an AI decision, signaling a trust breach." outputs: dial_change: direction: downgrade steps: 1 example: "notify → confirm, auto → notify" rebuild_protocol: description: "Re-runs the approval gate pattern for N interactions before re-proposing upgrade" n: configurable brain_update: layer: L2 signal: trust_breach_recorded effect: "Resets consecutive-approval counter for this domain" example: "Hana rejects a substitution she explicitly blocked. The system demotes from Notify → Suggest and confirms every swap until trust is re-earned." # ------------------------------------------------------------------- # A — Adaptation (8 patterns) # ------------------------------------------------------------------- - id: A1 name: Form Factor Transform category: A essence: "The same content restructures itself when the active device changes — a different vessel, not a different message." trigger_tokens: form_factor: any # Fires on any form_factor change trigger_extended: state_condition: form_factor_change: true description: "Fires whenever the primary display surface changes — phone to TV, phone to watch, etc." outputs: transformation_examples: phone_handheld_to_tv_display: "Scrollable list → full-bleed timeline with photos" phone_handheld_to_watch: "Full list → single next action only" phone_handheld_to_fridge_display: "Standard card → large touch target card with meal context" principle: "Content is the same. Presentation shape follows the vessel." example: "The family trip itinerary is a scrollable list on Hana's phone; cast to the TV it becomes a full-bleed timeline with photos." - id: A2 name: Cognitive Scaling category: A essence: "Information density and number of choices contract when Cognitive Load is high, expand when it drops." trigger_tokens: cognitive_load: [high, overloaded] trigger_extended: description: "Fires when cognitive_load is high or overloaded. Expansion fires when cognitive_load returns to low or moderate." outputs: scaling_table: low: { max_choices: 8, ui_density: high } moderate: { max_choices: 5, ui_density: medium } high: { max_choices: 3, ui_density: low } overloaded: { max_choices: 1, ui_density: low, defer_non_critical: true } example: "Running late at the airport, Hana sees three pre-selected options. At home on a quiet evening, she sees the full 12-item list." - id: A3 name: Social-Aware Filtering category: A essence: "Displayed content shifts based on who is present — prices, wish lists, and personal notes hide as exposure increases." trigger_tokens: social_exposure: [social_acquaintances, public, family_with_children] disclosure_dial: [none, minimal, moderate] # NOT full trigger_extended: description: "Fires when social_exposure is non-private AND disclosure_dial is not full." logic: "social_exposure >= social_acquaintances AND disclosure_dial != full" outputs: filtering_by_exposure: family_with_children: hidden: [gift_prices, adult_content, financial_details] social_acquaintances: hidden: [financial_information, health_data, message_previews] public: hidden: [all_personal_information] example: "Alone, Hana sees prices and budget stats. At a dinner party, the TV shows the recipe — no costs visible." - id: A4 name: Disclosure Cascade category: A essence: "Information visibility moves through four levels — FULL → SUMMARY → EXISTENCE → HIDDEN — per domain and per person." trigger_tokens: disclosure_dial: any # Any disclosure_dial value triggers this — the cascade IS the output trigger_extended: state_condition: social_exposure_change: true or: disclosure_dial_change: true description: "Fires on any change to disclosure_dial OR social_exposure that changes the effective visibility level." outputs: cascade_levels: full: "Complete information visible — Hana sees all of Koji's medical history" summary: "Category and status visible — doctor app sees 'has allergies: peanut'" existence: "Presence acknowledged, details hidden — school nurse form sees 'medical info on file'" hidden: "No information — strangers see nothing" example: "Koji's medical history: Full for Hana, Summary for the family doctor's app, Existence-only for the school nurse's form." - id: A5 name: Disposable Surface category: A essence: "A UI instance is born for a specific purpose and dissolves when that purpose is complete — tied to the Disposable Brain lifecycle." trigger_tokens: {} # No standard token trigger trigger_extended: intent_condition: intent_type: time_bound_goal description: "Fires when Intent is time-bound and scoped (trip, renovation, event)." brain_condition: brain_lifecycle: created brain_type: disposable outputs: lifecycle: [born, active, returning, dissolved] brain_coupling: born_with: "Disposable Brain instance for this goal" dissolves_when: "Goal time window closes" handoff_to: A8 # Temporal Handoff handles the learning return example: "The Kyoto trip UI appears the morning they leave, learns during the trip, and disappears when they arrive home." - id: A6 name: Care Architecture category: A essence: "AI tracks emotional and relational context to surface care opportunities buried in daily life." trigger_tokens: {} # No standard token trigger trigger_extended: brain_condition: layer: L2 signal: relationship_event_proximity description: "Fires when Brain L2 detects an upcoming relational event (birthday, anniversary, recurring care moment)." temporal_condition: proximity_window: configurable default: "7 days ahead" outputs: ui_pattern: ambient_care_suggestion must_not: "interrupt current task — always surface at Cognitive Load = low" cognitive_load_gate: true # Only surface when cognitive_load is low or moderate example: "The app surfaces 'Hana's mother's birthday is next week — last year you sent flowers' before she thinks to check." - id: A7 name: Live Recomposition category: A essence: "An existing plan is instantly rebuilt when external conditions change — weather, cancellations, or inventory — without the user seeing the seam." trigger_tokens: feasibility: [partially_feasible, infeasible] # Changed from feasible trigger_extended: state_condition: plan_state: active feasibility_change: external_trigger description: "Fires when an active plan's feasibility changes due to external event (weather, cancellation, stock)." note: "Unlike E3 Rollback, A7 operates before execution — the plan is rebuilt before the user needs to act on it." outputs: ui_pattern: silent_recomposition user_facing: "Updated plan shown as-is, without 'we changed this' friction" optional: "Contextual note for significant changes (cancelled venue vs minor time shift)" example: "The outdoor market is rained out. The itinerary silently swaps it for an indoor museum at the same time slot." - id: A8 name: Temporal Handoff category: A essence: "When a Disposable Brain ends, its distilled learnings return to the persistent Home Brain — the temporary context dissolves, the knowledge survives." trigger_tokens: {} # No standard token trigger trigger_extended: brain_condition: brain_lifecycle: ending brain_type: disposable description: "Fires when a Disposable Brain's lifecycle reaches its end state." outputs: handoff_process: - "Distill key learnings from Disposable Brain L2" - "Write to Home Brain L2 with source tag and date" - "Mark Disposable Brain as dissolved" written_to_home_brain: - "Preference discoveries (foods, pacing, comfort)" - "Behavioral patterns unique to that person in that context" - "Tradeoff decisions made under pressure" not_written: "Temporary operational data (booking confirmations, receipts)" coupling: A5 # Disposable Surface is created by A5; A8 handles its end example: "After the Kyoto trip, 'Kai loves ramen' and 'Mia fades at 3pm' are written into the Home Brain. The Trip Brain is gone." # ------------------------------------------------------------------- # X — Enterprise (4 patterns) # ------------------------------------------------------------------- - id: X1 name: Reasoning Trace category: X essence: "AI decision rationale is available in a collapsible format — auto-expands for high-risk decisions, silent otherwise." trigger_tokens: priority_weight: [high, critical] trigger_extended: user_action_condition: action: request_reasoning optional: true description: "Auto-expands for high/critical priority decisions. Also fires on explicit user request." outputs: ui_pattern: collapsible_trace auto_expand_at: [high, critical] trace_content: - data_sources_used - confidence_level - alternative_options_considered example: "The AI recommends Vendor A. Priya sees the answer; tapping 'Why?' reveals the three CRM threads and two calendar entries that drove it." - id: X2 name: Inline Edit category: X essence: "An AI proposal can be edited in place and immediately re-executed — the edit path within Approval Gate." trigger_tokens: autonomy_dial: confirm # Must be at Approval Gate to have an edit option trigger_extended: user_action_condition: action: select_edit context: "at a D1 Approval Gate" description: "Fires when user selects 'Edit' at a D1 Approval Gate, rather than Approve or Reject." outputs: ui_pattern: inline_editor on_save: "Re-execute AI process with edited input" coupling: D1 # X2 is the edit sub-pattern of D1 example: "The AI drafts a sales email. Priya changes one sentence inline; the system re-runs the draft and returns a revised version in seconds." - id: X3 name: Autonomy Dial UI category: X essence: "Explicit user control over the four-stage Autonomy level — Suggest, Confirm, Notify, Auto — paired with the Disclosure Dial." trigger_tokens: {} # Settings-invoked, not context-triggered trigger_extended: user_action_condition: action: explicit_dial_adjustment context: "settings or in-context dial control" description: "Fires when user explicitly adjusts Autonomy Dial. This is the UI pattern — not a reaction to a token state." outputs: ui_pattern: paired_dial_control requires_both: [autonomy_dial, disclosure_dial] note: "Always shows Autonomy and Disclosure together — the logical coupling must be visually represented." per_domain: true example: "Priya sets the brief-writing agent to Confirm for client-facing content, Notify for internal drafts, Auto for first-pass research summaries." - id: X4 name: Source Attribution category: X essence: "AI answers include the specific data sources that produced them — CRM, email, calendar — meeting audit and compliance requirements." trigger_tokens: priority_weight: [high, critical] # Compliance most critical for high-stakes decisions trigger_extended: compliance_condition: compliance_rule_active: true optional: true user_action_condition: action: request_source_evidence optional: true description: "Fires when compliance rules are active OR when user requests source evidence. Also auto-fires at high/critical priority." outputs: ui_pattern: source_chips chip_contents: "Data source label + tap-to-expand" required_for: "compliance-sensitive outputs" example: "'Q3 revenue up 12%' is shown with chips: CRM · Email · Calendar · Sheets — traceable in one tap." # ------------------------------------------------------------------- # Pattern Composition Map # ------------------------------------------------------------------- composition_notes: - id: silent_resolution name: "Silent Resolution (canonical composition example)" sequence: [D4, D6, E5, A2] description: > Omakase Mode (D4) runs autonomously until an action triggers Dynamic Friction (D6). If the user rejects the forced approval, Trust Breach Recovery (E5) demotes the dial. Meanwhile, Cognitive Scaling (A2) ensures the UI shows only critical info if load is high. source_page: "context-grammar/ax-patterns/index.html — 'Patterns Compose' section" - id: approval_with_inline_edit name: "Approval Gate + Inline Edit" sequence: [D1, X2] description: "D1 generates the approval UI; X2 is activated when the user selects the 'Edit' path." - id: disposable_brain_lifecycle name: "Disposable Brain lifecycle" sequence: [A5, A7, A8] description: "A5 creates the surface and brain. A7 adapts plans while active. A8 dissolves and extracts learnings." - id: trust_escalation_cycle name: "Trust escalation → breach → recovery cycle" sequence: [D2, D6, E5] description: "D2 builds trust and proposes dial upgrades. D6 maintains a safety ceiling. E5 handles breaches and resets."