Skip to content
← back

journey · doctrine evolution as a devlog arc

Eight anchors. Pain → action → result. Receipts attached.

AgentOps did not arrive as a finished doctrine — it accreted over five months of bookkeeping, public devlogs, and mixed-model councils that kept reframing the diagnosis. This page is the spine: eight inflection points where the thesis changed shape, each with the upstream artifact it points back to. Read it as a working changelog for the doctrine, not a launch narrative.

The eight anchors

  1. #01

    Bookkeeping Kernel

    Pain
    Platform team at JREN/NGA was re-deriving knowledge every time someone rotated off. Air-gapped enclaves, deep Helm chains, tribal knowledge as the only continuity layer — the kind of place a missed footnote becomes a 3 AM incident.
    Action
    Wrote it down. 163 how-tos, 138 reference docs, 29 onboarding tutorials. Bookkeeping as a first-class deliverable, not a cleanup task.
    Result
    The corpus stopped being a cost center and became the team's compounding asset. Bookkeeping became the kernel everything else grew from.

    The bookkeeping is the moat — agents make it nearly free, and the race is on.

    evidence: /home/boful/dev/agentops/docs/origin-story.mdsynced 2026-05-20
  2. #02

    Three Pillars

    Pain
    Three different problems — measuring human-AI sessions, operating agents in production, maintaining institutional memory — kept producing the same artifacts: planning rules, learnings, validation gates. Three frameworks competing for the same surface.
    Action
    Synthesized them into one architecture: Knowledge OS as the skeleton, Vibe Ecosystem measuring how knowledge compounded, 12-Factor AgentOps operationalizing the pipeline.
    Result
    Six laws (40% context, sub-agent isolation, git-native first, validation gates, human approval, learning extraction) became the spine that every downstream skill plugs into.

    Three frameworks, one architecture: measure compounding, operationalize the pipeline, govern with hard laws.

    evidence: /home/boful/dev/agentops/docs/origin-story.mdsynced 2026-05-20
  3. #03

    TSMC Foundry Thesis

    Pain
    Agent labs were racing on model quality. Teams shipping with agents were stuck — same model, wildly different outcomes. The differentiation wasn't the model; nobody had a name for what it actually was.
    Action
    Named it. Devlog 1 published the foundry thesis: the job is to build and run AI coding foundries; whoever engineers their knowledge compounding most deliberately wins. TSMC dominates semiconductors via yield, not machines.
    Result
    The foundry framing reset the question from "which model" to "what operational discipline around the model." It became the recruiting line for every later devlog.

    The job isn't to write code anymore — it's to build and run AI coding foundries.

    evidence: /home/boful/dev/agentops-showcase/docs/skeptics/not-vibe-coding.mdsynced 2026-05-20
  4. #04

    Open Source Bet

    Pain
    Internal methodology hardened in a private gitops repo. Hard to validate the doctrine without external pressure — easy to convince yourself a system "works" when only its author runs it.
    Action
    Open-sourced AgentOps. The three pillars became 33 CI gates; the six laws became 66 skills across four runtimes. Public from commit one.
    Result
    20 GitHub stars on launch day. The pressure of building in public forced rapid evolution — the product had to work for anyone who cloned it, not just its author. 1,083 commits later, the discipline held.

    Building in public is the validation gate methodology can't fake.

    evidence: /home/boful/dev/agentops-showcase/docs/case-studies/12factor-docs-site.mdsynced 2026-05-20
  5. #05

    Context Compiler Reframe

    Pain
    The methodology kept being described as "prompt engineering plus learnings," which understated it. Reviewers asked the same question: what is this, structurally? "A bunch of skills" is not an architecture.
    Action
    Reframed it as a compiler. Raw signal → curated learnings → compiled findings → planning rules and pre-mortem checks → enforcement gates that reject bad plans before implementation. The pipeline mirrors a traditional toolchain almost exactly.
    Result
    The compiler analogy unlocked the product story. Vendor memory follows the chat; the corpus follows the team. Context became an engineering artifact — typed, versioned, validated, decay-ranked — not chat history.

    AgentOps is the context compiler: compile context, gate output, compound knowledge.

    evidence: /home/boful/dev/agentops/PRODUCT.mdsynced 2026-05-20
  6. #06

    Mixed Model Sovereignty

    Pain
    The cross-vendor / mixed-model capability was easy to dismiss as marketing. Most multi-runtime tools test on the primary and shim everything else; if the second judge never disagrees in a load-bearing way, the second judge is decoration.
    Action
    Ran real /council --mixed sessions with independent Codex (gpt-5.5) and Claude judges. Stopped trying to prove the claim with benchmarks (v1 workbench was saturated) — proved it with in-repo artifacts where the cross-vendor judge moved the verdict.
    Result
    Three load-bearing findings the all-Claude councils missed, all with file:line citations. The /sovereignty-proof page made the claim falsifiable instead of rhetorical.

    Cross-vendor coordination is not decoration when the second vendor changes the outcome — and we have the receipts.

    evidence: /home/boful/dev/agentops/.agents/council/2026-05-15-rpi-leanness-council.mdsynced 2026-05-20
  7. #07

    RPI Leanness Reframe

    Pain
    RPI had become token-expensive and felt waterfall-like: big spec, big plan, then iterate. The instinct was to re-architect the methodology — agile-ify the loop, smaller waves, less ceremony.
    Action
    The mixed council reframed the diagnosis. Codex, surfacing as the independent voice, pointed at the real bug: plan.md was a shared god artifact, re-read ~8× per RPI lifecycle at 20–44 KB each time. Bloat is an artifact-boundary problem, not a methodology failure.
    Result
    execution-packet.json became the contract; plan.md became commentary. A mis-scoped re-architecture epic got cancelled. The fix was surgical, not structural.

    When something feels too heavy, audit the artifact boundary before re-architecting the methodology.

    evidence: /home/boful/dev/agentops/.agents/council/2026-05-15-rpi-leanness-council.mdsynced 2026-05-20
  8. #08

    CDLC And 3.0 Fungibility

    Pain
    The product kept being described in implementation terms — hooks, skills, packets — and reviewers couldn't place it in the existing software-engineering vocabulary. "What category is this?" had no clean answer.
    Action
    Named the lifecycle. Every SDLC phase has a context counterpart: the Context Development Life Cycle. Validation, compilers, code review, CI/CD, postmortems, runbooks — all translate directly. The 3.0 turn followed: fungibility-first agents, hookless CDLC rearchitecture, single-runtime default with mixed-model preserved as a sovereignty capability.
    Result
    CDLC gave the product a category. The 3.0 plan made install-to-first-value cheap without giving up the cross-vendor receipts. The doctrine arrived at its current shape: SDLC control plane, CDLC mechanism, fungible agents, sovereign councils on demand.

    The category is SDLC control plane for agentic software; the mechanism is CDLC; the agents are fungible by default and sovereign on demand.

    evidence: /home/boful/dev/agentops/docs/cdlc.mdsynced 2026-05-20

Source: agentops/docs/origin-story.md + agentops/PRODUCT.md + agentops/docs/cdlc.md + .agents/council/2026-05-15-rpi-leanness-council.md + .agents/council/2026-05-16-research-cdlc-claims.md · last synced: 2026-05-20

Provenance is checked by scripts/check-copy-freshness.mjs. Refresh anchors when upstream artifacts move.