Find your way around.

Navigate

Home: engineering bench and current work.

Navigate

Projects, proof, evidence, and open questions.

Navigate

First looks, opinions, lab notes, and teardowns.

Navigate

Tutorials, how-to procedures, and concepts.

Navigate

Comparisons and utilities.

Navigate

Reusable patterns, components, packs, and architectures.

Navigate

Documented interfaces and building blocks.

Navigate

Terms, glossary, and concepts.

Navigate

Source-pinned evaluations and workflows.

Navigate

The engineering bench and Tim.

Navigate

Participate, contribute, or start a conversation.

Library

Workflow graph, independent evaluator, verification, and policy boundaries.

Library / Components / draft

A bounded manifest that ties a proposed result to the checks, artifacts, provenance, and known gaps needed for review. agent-systems evals trust-boundary

Library / Components / draft

A scoped handoff that gives an accountable person the evidence, options, and authority needed to approve or stop a consequential action. agent-systems trust-boundary evals

Library / Components / draft

A separately controlled judge that compares a proposed result and its evidence with the original intent and a versioned rubric. agent-systems evals trust-boundary

Library / Components / draft

A versioned task contract that tells a software factory what outcome to pursue, what boundaries to respect, and how completion will be judged. agent-systems trust-boundary

Library / Components / draft

A separate decision point that allows, denies, or escalates a proposed action against versioned rules before the action takes effect. agent-systems trust-boundary runtime-infra

Library / Reference

Autonomic software engineering applies the closed control loop to the engineering process itself. Autonomic computing pointed MAPE-K at the runtime: the cluster manages itself. Autonomic software engineering points it at the repository: the codebase changes itself, observes the consequence, and converges against an eval rather than a reviewer. ASE

Library / Reference

The closure test is a four-condition rubric for whether an agent system actually closes the loop. Pass or fail, no maturity levels: closed loop, declarative target, unattended in the ordinary case with named exceptions, and ground truth — an eval decides whether the loop converged. Condition 4 is the disqualifier. closure closure rubric

Library / Reference

A reference architecture for autonomic software engineering that separates the agent that proposes a change (the Proposal plane) from the authority that decides whether the change helped (the Verdict plane). The Verdict plane is a separate component, with its own writer and its own version, that the Proposal plane submits to. Mechanical separation, not procedural. two plane loop two-plane architecture MAPE-K at the repository Verdict plane Proposal plane

Library / Packs

Closes the loop on agent evaluation by separating the Verdict plane from the Proposal plane, so a change to the corpus can never silently tune the judge that decides whether the change helped. autonomic-swe evals agent-systems hardened

Tools / Interactive model

Authority, permissions, guardrails, evidence, and approval. Reference only.

Tools

Coding agents, pricing, costs, subscriptions, Codex, Claude, and Grok.

Tools

An open-source library of canvas-rendered effects that mount over your live DOM without breaking selectability or click-through. runtime-infra agent-systems hardened

Notes / opinion

Tests tell an agent whether known checks passed; they do not automatically prove that the change helped users, respected every boundary, or is safe to keep.

Notes / opinion

Tool calls and busy run logs measure agent activity; progress begins when the workflow can prove that the target state moved closer and stayed there.

Notes / opinion

When an agent asks for approval too often and without decision-ready context, the workflow trains people to click through instead of protecting the actions that deserve judgment.

Notes / first look

More capable agents make it practical to delegate longer tasks, which makes clear outcomes and independent checks more useful than ever.

Notes / opinion

Decide how you will judge the result before building the agent, or your examples and tests will slowly start to favor whatever the agent already does well.

Notes / opinion

Semantic reasoning can inform policy decisions, but hard constraints, permissions, budgets, and operational invariants should remain outside the reasoning model wherever they can be deterministic. policy agent-systems trust-boundary authority-compiler

Notes / opinion

An agent should know how a consequential action will be contained or reversed before it acts, because a rollback invented after failure is only another untested production change.

Notes / lab note

Dynamic tool discovery is not only a context-efficiency optimization; it is an authority decision about which classes of action are currently available. agent-runtime permissions tool-design authority-compiler

Notes / opinion

When an agent changes a system faster than you preserve why its work was accepted, each successful run leaves future operators with uncertainty they will have to repay.

Notes / opinion

Let an agent explain and improve its work, but use separate evidence and authority for the decision to accept, publish, merge, or deploy it.

Notes / teardown

A general shell exposes ambient authority, while a narrow tool makes scope, auditability, and verification part of the interface. sandboxing tool-design agent-systems authority-compiler

Notes / teardown

Cursor can write code, run tests, and iterate, but a person still approves the work and the same agent that makes the change also interprets whether it worked.

Guides / concept / draft

AI operations IME is an emerging ReactorJet and Biro direction for turning operator intent into inspectable, bounded work without pretending that a prompt is authority or evidence. agent-systems trust-boundary

Guides / concept / draft

The control plane holds intent, policy, and authority while the work plane performs bounded actions and reports what happened. agent-systems runtime-infra trust-boundary

Guides / concept / draft

Agent engineering is controlled delegation of authority under uncertainty, with intent, discretion, constraints, and verification kept explicit. agent-systems trust-boundary evals

Guides / concept / draft

Confidence describes a belief about a result, while evidence records observations that another person or system can inspect and use to decide what happens next. evals agent-systems trust-boundary

Guides / concept / draft

A serious agent runtime decides which proposed transitions are admissible through policy, scoped capabilities, isolation, evidence, and recovery. agent-systems runtime-infra trust-boundary

Guides / concept / draft

Long-running agent work needs claims, evidence, freshness, and authority as live state rather than a final pass or fail checkpoint. evals agent-systems trust-boundary

Guides / concept / draft

Software is autonomic when it can observe the result of its own action, judge that result against a protected target, and choose the next bounded step. autonomic-swe agent-systems

Guides / how-to / draft

Place human approval at the smallest number of consequential transitions where judgment, accountability, or exception handling is still required. agent-systems trust-boundary

Guides / how-to / draft

Turn a written constraint into a separate, testable decision that permits or blocks a consequential agent action. agent-systems trust-boundary runtime-infra

Guides / how-to / draft

Separate evaluation assets and authority from the agent so a proposal cannot redefine the test that decides whether it passes. agent-systems evals trust-boundary

Guides / how-to / draft

Build an evidence bundle that connects each workflow claim to a reproducible observation, its source, and the decision that used it. agent-systems evals runtime-infra

Guides / tutorial / draft

Split preparation from execution so an agent can assemble consequential work while a person controls whether that exact action proceeds. agent-systems trust-boundary

Guides / tutorial / draft

Run an agent in a restricted workspace, submit its result to a separate checker, and keep only work supported by recorded evidence. agent-systems evals trust-boundary

Guides / tutorial / draft

Reconstruct a failed run from durable records, locate the first broken assumption, and choose a recovery without erasing the evidence. agent-systems runtime-infra evals

Work / working model

Deriving the capability surface available to an agent from delegation, policy, evidence, risk, and current task state. agent-systems trust-boundary runtime-infra How should evidence unlock temporary authority without allowing the reasoning model to redefine its own boundaries?

Work / working model

Exploring how the control-loop idea applies to software changes that observe consequences and converge against a protected target. autonomic-swe evals agent-systems What makes an engineering loop genuinely autonomic rather than an automation script with a model in the middle?

Work / exploring

A model for keeping claims, evidence, freshness, and next-action authority explicit across work that outlasts a single agent turn. evals agent-systems trust-boundary How should a long-running task represent what is proven, what is stale, and what remains safe to do next?

Work / working proof

Reference instantiation of the two-plane loop on a single repository. Repository repair per blueprint §6, blast radius L1 (Gated). autonomic-swe agent-systems runtime-infra What additional evidence would justify moving this bounded proof beyond L1 gated operation?

About

Works on the Verdict plane side of the two-plane loop. Designs scoring rubrics, curates evaluation corpora, and reviews closure-test scorecards. Background in formal verification and property-based testing; came to autonomic software engineering by way of asking "who checks the checker?" one too many times. verifier design and the closure test corpus curation and case authoring scoring rubric review

About

Works on the deployment side. Picks blast radii (L0 through L3) for live systems, designs the infrastructure reconciliation instantiation of the two-plane loop, and runs the cutovers when a system moves from L1 (Gated) to L2 (Unattended). Prefers boring solutions and finds them exciting; this is rarer than it sounds. two-plane loop deployment on real workloads blast radius selection and L2+ rollout infrastructure reconciliation projects

About

Platform engineer working at the intersection of platform engineering and increasingly capable AI systems. agent architecture applied AI product engineering platform architecture developer tooling