Glossary

159 key terms

Every key term the chapters define, with the chapter that defines it. Type to filter.

Agent defect classes: nonexistent APIs, phantom dependencies (imports of packages that do not exist or are not needed), tests that assert nothing, edited tests, plausible handling of ambiguity, and missing server-side checks. Chapter 12
Agent harness: the program that runs a coding agent, supplying tools, enforcing permissions, and looping until the task ends. Chapter 1
Agent-native rework: resolving a conflict by re-running the second change's mission against the merged code, so the agent writes a fresh implementation, instead of editing by hand (Hassan 2026, §6.4.5, p. 193). Chapter 11
Agent-sized task: a task satisfying one agent session, one reviewable change and one acceptance check. Chapter 11
Architecture decision record (ADR): a short repository document recording one decision: title, status, context, decision, consequences. Chapter 10
As-received baseline: the tagged state of an inherited project plus the characterization tests recording what it does. Chapter 18
Assumption audit: a six-step check of whether a method still works after its design conditions have changed. Chapter 2
Automatic translation (AT, ATPROD): percent of code translated by tool, and its productivity, priced apart from development. Chapter 6
BAC, PV, EV, AC, SPI, CPI, EAC: the quantities and indices of Section 7.1.4. Chapter 7
Baseline-scoped fit: a growth-model fit restricted to the period when program, model, prompts and configuration were unchanged. Chapter 15
Blackboard: agents read and write a shared store while a control component decides who acts. Chapter 8
Bound property: an acceptance property linked to the check that tests it. Chapter 12
Brooks's law: adding people to a late software project makes it later. Chapter 6
Causal provenance: the record tying a change to the mission and process configuration behind it. Chapter 14
Change footprint: the number of modules read or edited to make one change correctly. Chapter 10
Change rate: the share of approvals at a touchpoint after which the human changed or rejected the work. Chapter 3
Characterization test: a test that pins what code currently does, without claiming it is correct. Chapter 11
Clarification gate: a workflow step requiring the agent to raise open questions before implementing, and blocking work while open decisions are unresolved. Chapter 4
Classical bound: the 75% ceiling for players who share only what is fixed in advance and do not communicate. Chapter 17
COCOMO II: the Constructive Cost Model, revised 2000; effort and schedule from size, scale factors and effort multipliers. Chapter 6
Code-complete, merge-ready, integrated: an agent reports done; reviewed and acceptance check passed; merged and main line green. (Hassan's third rung is "integration-ready": Hassan 2026, §6.4.4, p. 192.) Chapter 7
Code-generation cycle: the loop from a specification to code and its checks. Chapter 3
Coding agent: a model given tools to read and edit files and run commands. Chapter 1
Compliance by declaration: accepting that a check happened because the actor says so (Hassan 2026, §9.9.6, p. 277). Chapter 16
Conditional likelihood: the likelihood that an intentional threat succeeds given that the hostile agent is present (NASA 2022). Chapter 9
Cone of uncertainty: the narrowing range of estimate error as a project progresses. Chapter 6
Configuration drift: configuration in use that differs from the recorded baseline. Chapter 14
Confused deputy: a program that uses its own authority for a caller's purpose (Hardy 1988). Chapter 9
Congruence: the match between the dependencies that exist and the channels that carry them. Chapter 8
Consistency without correctness: agreement among reviewers that share blind spots. Chapter 12
Context budget: a written allocation of an agent's working space for a task into fixed, task and headroom parts. Chapter 11
Convention: a default, fixed before the round, that separated parties adopt because each expects the other to. Chapter 17
Convention file: a repository file (AGENTS.md, CLAUDE.md) holding project guidance that an agent loads at the start of a session. Chapter 10
Converge-or-split decision: a two-agent decision with private bits where success means matching actions unless both bits are unusual. Chapter 17
Conway's law: a system's structure tends to copy the communication structure of its builders (Conway 1968). Chapter 1
Coordination: the work of making the decisions of many workers fit together in one system. Chapter 1
Coordination failure: a defect that lives between parts, where each part satisfies its builder's understanding and the understandings differ. Chapter 1
COSMIC Function Point (CFP): one data movement of one data group (ISO/IEC 19761). Chapter 5
Cost-of-change curve: the claim that the cost of fixing a defect or decision rises the longer it persists. Chapter 3
Coverage: the share of code a suite executes; not a measure of adequacy. Chapter 13
Credit event: the stated event at which a work package earns its budget. Chapter 7
Decision gate / execution gate: a human choice among alternatives the pipeline cannot rank; a human approval of work the pipeline is doing as designed. Chapter 3
Decision rule: the ordered questions Q1 to Q4 that assign a verdict (Section 2.5). Chapter 2
DecisionBOM: Hassan's record of how a change was decided. Chapter 14
Deep module: a module with a simple interface in front of substantial hidden functionality. Chapter 10
Defect-encoding test: a test whose expected value contradicts the specification. Chapter 13
Defensive programming: writing code to protect itself from invalid inputs and its own errors, through validation, assertions and barricades. Chapter 11
Delivered-product configuration: the code, tests, build configuration and lockfile that ship. Chapter 14
Delta-first review: reviewing only what changed since the last approval (Hassan 2026, §7.3.3, pp. 221–222). Chapter 12
Demonstrated customer value: value shown in use after release, as distinct from budgeted value earned. Chapter 7
Do Not section: the part of a convention file listing prohibitions, each with a reason and an enforcement pointer. Chapter 10
DRY: every piece of knowledge has one authoritative representation within a system. Chapter 10
EARS: Easy Approach to Requirements Syntax; five sentence templates. Chapter 4
Effective leak fraction (q): the share of rounds in which a leak was present and used correctly; win rate ≈ 0.75 + 0.25 × q. Chapter 17
Engineering manager of an agent team: your role: you specify, decompose, assign, review, verify, decide, escalate and record, and answer for the result. Chapter 18
Equivalent size (ESLOC): adapted code converted to equivalent new lines through AAM, built from DM, CM, IM, AA, SU and UNFM. Chapter 6
Error detection efficiency: errors found by an inspection divided by total errors in the product before it, times 100. Chapter 12
Escalation matrix: decision type, owner, consult path and required evidence. Chapter 8
Eval: a test suite for a nondeterministic component: inputs, a scoring rule, repeated trials, a rate. Chapter 13
Evidence chain: observation, inference, proposed mechanism, test. Chapter 2
Evidence package: the six-part set of artifacts, each produced by someone other than the author or marked as the author's statement, that an inspector audits instead of the raw diff. Chapter 12
Executable specification: a specification whose every statement of required behavior is a runnable check or bound to one. Chapter 4
Expanded baseline: a baseline that also lists generation-process configuration and eval results. Chapter 14
Fagan inspection: a formal review with a moderator, defined roles, preparation, a meeting that finds but does not fix errors, rework and follow-up. Chapter 12
Fault, error, failure: a flaw in code or design; the incorrect internal state it causes; the visible departure from required behavior. Chapter 13
Functional size / function point (FP): size measured by the functions users require (ISO/IEC 14143-1); IFPUG's unit, counted as unadjusted function points (UFP). Chapter 5
Gearing factor / backfiring: average source lines per function point for a language; applied in reverse to estimate function points from code. Chapter 5
Generation ratio (γ): first-pass lines presented for review per line delivered, each counted once. Chapter 6
Generation-process configuration: the model, harness, prompts, convention files, skills and permissions that steer the agent writing changes. Chapter 14
Gherkin: the Given/When/Then format for scenarios. Chapter 4
Goel–Okumoto model: nonhomogeneous Poisson growth model, m(t) = a(1 − e^(−bt)); a is expected total faults, b the per-fault detection rate. Chapter 15
Hallucinated dependency: a package or interface a model recommends that does not exist. Chapter 9
Handoff package: the specification and backlog, convention file, decision records, risk register, work log and open issues that travel with a project, plus the baseline record and characterization-test pointers. Chapter 18
Handoff packet: a structured bundle of work artifact, evidence bundle, context summary, task specification and handoff metadata (Hassan 2026, §6.6.4, p. 201). Chapter 8
Handoff rubric: accuracy, specificity and usefulness, each scored 0 to 2 per package item. Chapter 18
Held-out test: predicting a project the model was not fitted to; a refit that covers the actual shows only compatibility. Chapter 18
Hidden coupling: a dependence between workers' outputs that no declared interface and no logged channel carries. Chapter 8
Holds, recalibrate, breaks, inverts: the assumption survives; new values or definitions restore the method; the method needs new variables, records or steps; an activity's scarce and abundant resources have traded places. Chapter 2
Implementation separation, diversity, independence: running layers as separate sessions or prompts; differing in model, method or information; a measured joint miss rate equal to the product of the miss rates. Chapter 16
Incremental detection: the share of one layer's misses that another layer catches (Chapter 16, condition 1). Chapter 15
Incremental detection: the share of the generator's missed defects that a checking role catches. Chapter 16
Information hiding: designing each module to conceal one design decision from the rest of the system (Parnas 1972). Chapter 8
Joint miss rate, φ: the fraction of defects missed by every layer; the correlation between two layers' miss indicators. Chapter 15
Leak test: dealing fresh fair bits to agents in a real pipeline and comparing the team's win rate with 0.75. Chapter 17
Leave-one-project-out: estimating each project from a model fitted to the others. Chapter 6
Legacy code: in Feathers's sense, code without tests. Chapter 11
Lethal trifecta: access to private data, exposure to untrusted content and the ability to communicate externally, in one agent (Willison 2025). Chapter 9
Levels of "done": code-complete, merge-ready, and integrated (this chapter's lab; Chapter 7). Chapter 1
Lines generated / presented for review / delivered: all lines an agent produces; lines first submitted to a reviewer; lines merged and kept. Chapter 5
Lines-of-code paradox: more expressive languages show lower lines per staff-month because fixed costs spread over fewer lines. Chapter 5
LLM-as-judge: using a model to grade another model's output; subject to position, verbosity and self-preference biases. Chapter 13
Load-bearing assumption: an assumption about a method's setting whose failure would make the method's output unreliable (term from Dewar 2002). Chapter 2
Loop / loop length: one pass from decision to evidence, and its calendar time. Chapter 3
Magnitude of relative error / PRED(25): the error of one estimate; the share of estimates within 25%. Chapter 6
Managing complexity: McConnell's primary technical imperative: organizing a program so that a person can focus on one part at a time. Chapter 11
Message passing: agents exchange explicit messages in agreed formats; the pipeline is the common case. Chapter 8
Mutant, equivalent mutant: a copy of the program with one small change; one that behaves identically and cannot be killed. Chapter 13
Mutation score: mutants killed divided by mutants generated minus equivalent mutants. Chapter 13
Naive baseline: COCOMO II at default settings with no local calibration. Chapter 6
Negation test: negating a candidate assumption to see whether the method's output becomes unreliable. Chapter 2
Non-change manifest: a list of what was deliberately left unchanged and why (Hassan's "Changes and Non-Changes Manifest"). Chapter 12
Non-goal: an explicit statement of what is out of scope. Chapter 4
Open decision: a stakeholder's choice not yet made, recorded with an owner and a default. Chapter 4
Operational profile: the operations a system performs and the probability of each in use (Musa 1993). Chapter 15
Orchestrator–worker: a lead agent decomposes, delegates and integrates; workers talk only to the lead. Chapter 8
pass@k, pass^k: the probability that at least one, or all, of k trials succeed; pass^k equals pᵏ when trials are independent (Yao et al. 2024; Chapter 15). Chapter 13
Phase boundary: what must be settled before the next kind of work begins. Chapter 3
Positive control: a planted leak that confirms the test can detect a known channel. Chapter 17
Pre-registration: writing down a prediction or estimate, dated and unchanged, before the data exist. Chapter 1
Predict-then-run: writing down the expected behavior diff before running the tests, so a wrong guess measures a comprehension gap. Chapter 11
Probe task: a small change where the model's habit and the written convention disagree, used to measure compliance. Chapter 10
Prompt injection: an attack in which instructions planted in content a model processes are followed as if legitimate; indirect when planted in content the agent later reads. Chapter 9
Property Coverage Index (PCI): the fraction of acceptance properties bound to a named check and evidence location. Chapter 4
Property-based acceptance: acceptance stated as properties that hold over a domain of inputs. Chapter 4
Protected path: a file agents may propose changes to but never merge. Chapter 14
Provenance trailer: a commit-message line recording model, mission, baseline, autonomy, evidence, director and approver. Chapter 14
Read-only acceptance test: a spec-derived test the implementing agent cannot modify, enforced by permissions and CI. Chapter 13
Read-time ratio: observed time to approve divided by the time reading would take at your review rate. Chapter 3
Recertification: rerunning the qualification and eval suites after any change to an item that steers the agents. Chapter 14
Recordable, not restorable: an item whose identifier you can note and whose change you can detect, but whose state you cannot recreate on demand. Chapter 14
Release cycle: the loop from a merged change to use in production. Chapter 3
Reversibility / blast radius: whether an action's effects can be undone at tolerable cost (R1 to R3) / what one failure can affect before a person can intervene (B1 to B3). Chapter 9
Review capacity: the hours per week reviewers can give (h), and what they can verify in a loop (h × T / k). Chapter 3
Review rate (r): lines reviewed carefully per hour; 400 is this book's illustrative figure, the fast end of verified rates. Chapter 6
Review-state gap: packages generated but not yet integrated. Chapter 7
Rework fraction (ρ): re-review, re-verification and re-prompting after defects, as a share of first-pass effort. Chapter 6
Risk exposure (RE): probability of an unsatisfactory outcome times the loss if it occurs. Chapter 9
Risk reduction leverage (RRL): reduction in exposure per unit of cost. Chapter 9
Role card, escalation matrix: a repository file giving a role's owner, model, permissions and qualification record; a table of decision type, owner, consult path and required evidence. Chapter 16
Role owner / decision owner: who answers for a role's configuration, and who answers for the decision it informs. Chapter 16
Scale exponent (E): the power on size in COCOMO II; E > 1 means diseconomy of scale. Chapter 6
Shared context: everything agents hold in common before a task starts. Chapter 8
Shared message: information one agent produces during the work and sends to another. Chapter 8
Signpost: an observable event that would warn that a vulnerable assumption is failing (Dewar 2002). Chapter 2
Single agent with full context: one agent holds the whole task; no inter-agent channels. Chapter 8
Size of the verification job: the pair (R, P) of first-pass lines presented and acceptance properties to be checked. Chapter 5
Software crisis: the disputed label some 1968 NATO participants gave to the gap between what large systems demanded and what methods delivered. Chapter 1
Specification package: the specification plus its linked, versioned decisions and context. Chapter 4
Stakeholder-feedback cycle: the loop from a working increment to a decision by the people it is for. Chapter 3
Statement of practice: a description of what practitioners do and why; without documented cases or measurements, not validation. Chapter 2
Steering (H_steer): prompting, hand-coding and debugging per agent task, t_s × n. Chapter 6
Stricter game: a game whose accepted action pairs are a subset of another's. Chapter 17
Structural exposure: risk that follows from an agent path's wiring, assessed without an unconditional probability. Chapter 9
Surprise log: the agent's record of anything unexpected it found; a lead for the inspector. Chapter 12
Tacit knowledge: knowledge people hold and use but cannot fully state (Polanyi 1966). Chapter 4
Test independence: the expected value in a test does not derive from the code under test. Different authorship alone does not establish it. Chapter 13
Test oracle: the mechanism that decides whether observed behavior is correct. Chapter 13
The reveal: the scoring of pre-registered estimates against actuals at the end of the term. Chapter 18
Three-arm inspection: the lab comparing raw-diff inspection, evidence-package inspection and an AI reviewer against ground truth. Chapter 12
Two-part register: Part A, probabilistic project risks; Part B, structural adversarial exposure. Chapter 9
Uncounted channel: a path carrying information about one agent's task to the other during a round, not listed in the design as communication. Chapter 17
Unit of analysis: the specific mechanism or activity an audit examines. Chapter 2
Uses relation: A uses B if A's correct functioning depends on a correct B; a loop-free relation yields subsets. Chapter 10
Verification debt: the gap between code generated and code verified (Hassan 2026, §10.1, pp. 289–290). Chapter 6
Verification load (k): hours of review and verification, with rework, per delivered line. Chapter 3
Verified merge: arrival of the integrated state; the credit event used for Dispatch. Chapter 7
WBS, work package, 100% rule: a decomposition of scope; its lowest element; child elements account for exactly 100% of their parent (Haugan 2002). Chapter 7
Wilson interval, rule of three: an interval for a proportion with good coverage at small n; with no failures in n trials, the 95% upper bound on the failure probability is about 3/n. Chapter 15
φ (phi): the share of a package's budgeted effort that remains after the point where it is credited. Chapter 7