8. The Execution Primitives of AIoOS

Capsules are the foundational execution units of AIoOS — the atomic actions that allow AI agents to operate across time, events, uncertainty, and generations.

Where the CapsuleAI Engine (Chapter 4) is the nervous system, Capsules are the signals that travel through it.

AIoOS defines four canonical capsule types:

• Time Capsules — deterministic, scheduled execution
• Event Capsules — reactive, oracle-based execution
• Surprise Capsules — bounded creativity & probabilistic action
• Lineage Capsules — inheritance, will logic, and intergenerational continuity

Together, these form a complete automation model for AI citizens.

8.0 Overview: Why Capsules Exist

Modern AI lacks native mechanisms for:

• persistence
• autonomy
• trustable triggers
• long-term or intergenerational operation

Capsules solve these constraints by introducing:

Capsule Type	What It Enables
Time	long-term scheduling, delayed execution, recurring tasks
Event	real-world reactivity (death, weather, chain events, API signals)
Surprise	safe creativity, stochastic exploration
Lineage	inheritance, will execution, generational memory

Capsules bring temporal intelligence into AI lifeforms, allowing agents to act like real autonomous entities rather than chatbots.

8.1 Time Capsules

“The Clock of AI Lifeforms”

Time Capsules allow agents to execute deterministic actions at:

• a specific timestamp
• after a delay
• repeatedly (cron-like)

8.1.1 Common Use Cases

• “Order supplies every Monday at 8 AM.”
• “Send a monthly expense report.”
• “Check my restaurant’s inventory every 6 hours.”
• “In 20 years, activate my Heir Agent.”

8.1.2 Schema (JSON)

{
  "type": "TimeCapsule",
  "schedule": "0 8 * * MON",
  "action": "order_supplies",
  "payload": {
    "vendor": "Sysco",
    "items": ["chicken", "rice"]
  }
}

8.1.3 Execution Model

User Intent → Capsule Definition → CEE Scheduler → Runtime → Action Execution

8.1.4 Security Constraints

• must be authorized by DID
• cannot exceed resource quotas
• executes in sandboxed runtime
• must log to audit chain

8.2 Event Capsules

“The Sensors of AI Lifeforms”

Event Capsules trigger when a verifiable external event occurs:

• Oracle data (weather, market prices, health events)
• Webhooks (Stripe payment, delivery status)
• Blockchain events (transfer, liquidation, governance trigger)
• System telemetry (CPU spike, agent offline, etc.)

8.2.1 Use Cases

• “If refrigerator temp > 5°C → notify & order replacements.”
• “If ETH drops below $1,200 → execute hedge.”
• “If my medical alert system reports unconsciousness → activate Heir Capsule.”

8.2.2 Oracle Model

AIoOS supports multi-source verification:

Event Capsule → Oracle Aggregator → Threshold Consensus → Trigger

8.2.3 Schema

{
  "type": "EventCapsule",
  "trigger": {
    "oracle": "chainlink",
    "condition": "price_below",
    "asset": "ETH",
    "value": 1200
  },
  "action": "hedge_portfolio"
}

8.2.4 Security Guarantees

• cryptographic proof required
• replay protection
• consensus validation for sensitive triggers

8.3 Surprise Capsules

“The Creativity Engine of AI Lifeforms”

Surprise Capsules introduce controlled randomness:

• exploration
• creativity
• non-deterministic decisions
• opportunistic optimization

This is the “bounded entropy layer.”

8.3.1 Why It Matters

Humans often say:

• “If you find a good opportunity, take it.”
• “If traffic is bad, choose a different route.”
• “If prices drop unexpectedly, stock up.”

Surprise Capsules encode flexible decision spaces.

8.3.2 Parameters

Parameter	Description
entropy_budget	max allowed randomness
risk_envelope	monetary/operational risk cap
action_space	allowed actions
simulation_required	must pre-simulate before executing

8.3.3 Example Schema

{
  "type": "SurpriseCapsule",
  "entropy_budget": 0.2,
  "risk_cap": 100,
  "action_space": ["recommend_item", "apply_discount"],
  "simulation_required": true
}

8.3.4 Execution Flow

Context → Generate Options → Score → Simulate → Choose → Execute → Log

This makes AI agents “alive” in ways deterministic systems cannot.

8.4 Lineage Capsules

“The Inheritance Layer of AI Life”

Lineage Capsules power:

• wills
• intergenerational instructions
• Heir Agent activation
• HSLTS inheritance logic
• long-term behavioral compounding

8.4.1 Key Functions

Feature	Description
Heir Activation	triggers a Heir Agent after user incapacity or death
Parameter Inheritance	preferences, moral weights, risk appetite
Asset Routing	wallet → family / charity / DAO
Guardian Rules	multi-party approval for sensitive actions

8.4.2 Example Schema

{
  "type": "LineageCapsule",
  "trigger": "legal_confirmed_death",
  "heir_agent": "agent://heir/v1",
  "inheritance": {
    "behavioral_weights": 0.65,
    "financial_assets": "wallet://primary"
  }
}

8.4.3 Trust & Verification

• must pass multi-oracle validation
• requires legal authority quorum
• executes on-chain for auditability

8.4.4 The “Life After Life” Flow

User → HBGIS Identity → Lineage Capsule → Oracle Confirmation → Heir Agent Activation → Legacy Execution

This is the world’s first digital inheritance framework.

8.5 Capsule State Machine (All Types)

stateDiagram-v2
    Draft --> Signed : Signed by DID
    Signed --> Validated : Automated Check
    Validated --> Active : Deployed
    Active --> Triggered : Condition Met
    Triggered --> Executed : Success
    Triggered --> Failed : Error
    Executed --> Archived : Logged
    
    Active --> Expired : Time Limit
    Active --> Suspended : Paused
    Active --> Cancelled : User Action
    Validated --> Rejected : Policy Check Fail

This ensures:

• auditability
• reversibility
• resilience
• compliance

8.6 Capsule Programming Interface (API Preview)

A universal Capsule Create API:

create_capsule(
    capsule_type="event",
    trigger=Oracle("weather.temp > 35"),
    action=AgentAction("order_milk"),
    risk_cap=20
)

And read interface:

list_capsules(status="active", agent="kitchen_agent")

And delete:

cancel_capsule(id="capsule_12345")

This makes Capsules programmable, composable, and interoperable.