Practical notes: Your AI Agent Will Not Survive Article 12

This walkthrough rebuilds the path from raw materials to a working system for: Your AI Agent Will Not Survive Article 12. The focus is operable steps, explicit checks, and code that you can drop into a repo without guessing intent. For the Overview stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

The cheap version of this already went to court

When working through the The cheap version of stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Checkpoint after expensive steps. Resume should not re-bill the same LLM call when an operator retries a later node.

Debugging is not accounting

When working through the Debugging is not accounting stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Checkpoint after expensive steps. Resume should not re-bill the same LLM call when an operator retries a later node.

The 3am version of this problem

When working through the The 3am version of stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion. Checkpoint after expensive steps. Resume should not re-bill the same LLM call when an operator retries a later node. When working through the The 3am version of stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

Better memory will not save you

The Better memory will not stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Keep graph state flat and typed. Nested blobs hide which node wrote which field and break resume after interrupts.

Make the decision a real object

The Make the decision a stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Keep graph state flat and typed. Nested blobs hide which node wrote which field and break resume after interrupts.

from semantica.context import ContextGraph

ctx = ContextGraph(advanced_analytics=True)
decision_id = ctx.record_decision(
    category="incident_triage",
    scenario="Checkout 5xx elevated 20 minutes after release 2026.31",
    reasoning="Error onset correlates with rollout window; no infra alerts; "
              "no dependency alarms visible at time of assessment",
    outcome="classified_as_release_regression",
    confidence=0.83,
)

One record is a log line. A chain is infrastructure.

The One record is a stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion. Keep graph state flat and typed. Nested blobs hide which node wrote which field and break resume after interrupts. The One record is a stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

triage_id = ctx.record_decision(
    category="incident_triage",
    scenario="Checkout 5xx elevated after release 2026.31",
    reasoning="Error onset correlates with rollout window; no infra alerts",
    outcome="classified_as_release_regression",
    confidence=0.83,
)

rollback_id = ctx.record_decision(
    category="remediation",
    scenario="Contain checkout 5xx",
    reasoning="Regression attributed to release; rollback is lowest-risk action",
    outcome="rolled_back_to_2026.30",
    confidence=0.91,
)
unblock_id = ctx.record_decision(
    category="release_gate",
    scenario="Resume deploy queue after containment",
    reasoning="Error rate normal for 30 minutes post-rollback",
    outcome="deploy_queue_reopened",
    confidence=0.86,
)
ctx.add_causal_relationship(triage_id, rollback_id, relationship_type="CAUSED")
ctx.add_causal_relationship(rollback_id, unblock_id, relationship_type="INFLUENCED")
why = ctx.trace_decision_chain(unblock_id)
blast_radius = ctx.analyze_decision_impact(triage_id)

Causality is not provenance

For the Causality is not provenance stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Put human approval on edges that spend money or change production data. Compile-time wiring does not equal business completeness.

from semantica.provenance import ProvenanceManager

prov = ProvenanceManager(storage_path="./provenance.db")
prov.track_entity(
    entity_id="release_2026_31",
    source="ci/pipeline_runs/2026-31/manifest.json",
    metadata={"stage": "canary", "extractor": "ManifestParser", "confidence": 0.99},
)
lineage = prov.get_lineage("release_2026_31")

prov.invalidate(
    "release_2026_31",
    agent_id="human_sre_lead",
    reason="Error attribution corrected; degradation traced to payments provider",
)

Get the policy out of the prompt

For the Get the policy out stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call.

from semantica.reasoning import ReteEngine, Rule, Fact, RuleType

engine = ReteEngine()
engine.build_network([
    Rule(
        rule_id="dual_control_prod",
        name="Human approval required for high blast-radius prod changes",
        conditions=[
            {"field": "env", "operator": "==", "value": "production"},
            {"field": "blast_radius", "operator": "in", "value": ["high", "critical"]},
        ],
        conclusion="require_human_approval",
        rule_type=RuleType.IMPLICATION,
    )
])
engine.add_fact(Fact("chg_014", "change", [{"env": "production", "blast_radius": "high"}]))
matches = engine.match_patterns()

Boring standards outlive your framework

For the Boring standards outlive your stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion. Put human approval on edges that spend money or change production data. Compile-time wiring does not equal business completeness. For the Boring standards outlive your stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

What this actually costs you

When working through the What this actually costs stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Checkpoint after expensive steps. Resume should not re-bill the same LLM call when an operator retries a later node.

Start with one workflow

When working through the Start with one workflow stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Checkpoint after expensive steps. Resume should not re-bill the same LLM call when an operator retries a later node.

The thing worth remembering

When working through the The thing worth remembering stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion. Checkpoint after expensive steps. Resume should not re-bill the same LLM call when an operator retries a later node. When working through the The thing worth remembering stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

Operational checklist

The Operational checklist stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope.

Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments.

Keep graph state flat and typed. Nested blobs hide which node wrote which field and break resume after interrupts.

Add a smoke test that exercises the critical path in CI with fixtures, not live paid APIs, whenever budgets allow.

Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

Keep graph state flat and typed. Nested blobs hide which node wrote which field and break resume after interrupts.

Before promoting the stack, freeze versions, capture a golden transcript for the critical path, and confirm rollback steps. Shared environments need rate limits, tenancy checks, and a clear owner for secret rotation. Prefer boring reliability over clever one-off demos.

Batch note for ae6c908d5bef: keep provider keys out of the repo, set a per-session token ceiling, and store transcripts next to the eval fixtures so later model swaps stay comparable.

For the hardening note 0 stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline.

Hardening detail 0/966: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

When working through the hardening note 1 stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments.

Hardening detail 1/966: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

The hardening note 2 stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish.

Hardening detail 2/966: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

For the hardening note 3 stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion.

Hardening detail 3/966: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

When working through the hardening note 4 stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

Hardening detail 4/966: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

The hardening note 5 stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline.

Hardening detail 5/966: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.

For the hardening note 6 stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments.

Hardening detail 6/966: measure wall time, error class, and token spend for this note, then decide whether to keep the change based on a fixed question set rather than anecdote.