Skip to content
Rig Docs

The Trusted Rig — Whitepaper

The architectural specification for the engineering rig we would trust to complete architect-level work, ship autonomously within bounded blast radius, and fix its own bugs with near-zero downtime.

This is a target document, not a description of the rig as it exists today. Closing the gap is a roughly six-month roadmap. For current state, see Implementation Status.

Core chapters

Section titled “Core chapters”

Reading order follows the MkDocs ToC. Each chapter is self-contained; the index is the reading guide.

#ChapterWhat it covers
1Design PrinciplesTen rules every other document is answerable to
2Trust ModelWhat the rig decides alone; tiered autonomy by blast radius
3SafetyGuards, stuck detection, hallucination mitigation, prompt injection defense
4SecuritySupply chain (Sigstore + SLSA), runtime (Kyverno + Cilium), attestation, secrets
5ObservabilityOpenTelemetry + Langfuse/Phoenix + Prometheus, SLOs, traces, cost attribution
6Cost FrameworkLiteLLM proxy, per-agent budgets, rate-limit-aware dispatch, prompt caching
7Self-HealingFlagger canary, SLO-gated rollout, kill switches, DB migrations
8Quality & EvaluationNightly eval harness, SWE-bench Pro, property-based testing, DORA metrics
9Drift DetectionModel drift, prompt drift, code drift, config drift
10Memorypgvector-backed agent memory: schema, 5 MCP tools, limitations
11Implementation StatusSingle source of truth for what’s deployed vs. planned vs. deferred
12MVP ScopeMinimum viable rig: 10 capabilities, ~3-4 pair-mode weeks

Supplementary chapters

Section titled “Supplementary chapters”
ChapterWhat it covers
Planner-EConversational intake and decomposition; Discord-thread-driven planning that converges on epics + child issues with TaskSpecs and tier classification
Tool ChoicesADR-style evaluation of every tool picked
Provider PortabilityVendor-neutral at four layers; Claude is default, others work too
LimitationsWhat the rig cannot do; where humans remain indispensable
GlossaryVocabulary for both humans and AI readers

Architecture overview

Section titled “Architecture overview”

The whitepaper’s ten properties in one diagram:

graph TB
    classDef foundation fill:#e8f5e9,stroke:#2e7d32,color:#000
    classDef safety fill:#fff3e0,stroke:#e65100,color:#000
    classDef autonomy fill:#e3f2fd,stroke:#1565c0,color:#000
    classDef honesty fill:#fce4ec,stroke:#ad1457,color:#000

    P1[1. Measurable]:::foundation
    P2[2. Bounded blast radius]:::safety
    P3[3. Reversible before irreversible]:::safety
    P4[4. Execute, don't trust]:::foundation
    P5[5. Attestable + replayable]:::foundation
    P6[6. Progressive autonomy]:::autonomy
    P7[7. Humans at semantic boundaries]:::autonomy
    P8[8. Trusted control + untrusted data]:::safety
    P9[9. Fail closed, fail known]:::honesty
    P10[10. Simple enough to operate]:::honesty

    P1 --> P6
    P4 --> P2
    P5 --> P6
    P8 --> P2
    P2 --> P3
    P9 --> P7
    P6 --> P7
View Mermaid source
graph TB
    classDef foundation fill:#e8f5e9,stroke:#2e7d32,color:#000
    classDef safety fill:#fff3e0,stroke:#e65100,color:#000
    classDef autonomy fill:#e3f2fd,stroke:#1565c0,color:#000
    classDef honesty fill:#fce4ec,stroke:#ad1457,color:#000


    P1[1. Measurable]:::foundation
    P2[2. Bounded blast radius]:::safety
    P3[3. Reversible before irreversible]:::safety
    P4[4. Execute, don't trust]:::foundation
    P5[5. Attestable + replayable]:::foundation
    P6[6. Progressive autonomy]:::autonomy
    P7[7. Humans at semantic boundaries]:::autonomy
    P8[8. Trusted control + untrusted data]:::safety
    P9[9. Fail closed, fail known]:::honesty
    P10[10. Simple enough to operate]:::honesty


    P1 --> P6
    P4 --> P2
    P5 --> P6
    P8 --> P2
    P2 --> P3
    P9 --> P7
    P6 --> P7
Section titled “Related”