# Agent Concepts

> Structural-reasoning counterpart to Agent Skills — a catalog of recurring relational concepts that show up across many independent domains.

## Docs

- [active-gate-vs-passive-audit](https://agentconcepts.io/concepts/active-gate-vs-passive-audit.md): Posture-toward-error: synchronous gate (block invalid input at the boundary) vs asynchronous audit (record, surface on schedule, respond after).
- [adapter](https://agentconcepts.io/concepts/adapter.md): Translates between two incompatible interfaces so a client can use a service whose shape it cannot natively consume; the adapter lives at the seam and absorbs the mismatch.
- [anchoring](https://agentconcepts.io/concepts/anchoring.md): An initial reference value disproportionately influences subsequent estimates, even when the anchor is irrelevant to the judgment at hand; bias on point-estimate magnitude rather than emergent topology.
- [arbitrage](https://agentconcepts.io/concepts/arbitrage.md): Exploiting a price (or value) discrepancy for the same good across two markets, capturing the spread until the act of exploitation closes it. The structural feature is the simultaneous existence of differently-priced replicas plus a frictional channel between them that an arbitrageur can profitably…
- [arms-race](https://agentconcepts.io/concepts/arms-race.md): Two adversaries escalate in coupled lockstep, each party's investment answered by a matching counter-investment, so capability on both sides ratchets monotonically upward without either gaining lasting advantage. The defining signature is escalation, not oscillation.
- [asymmetric-gate](https://agentconcepts.io/concepts/asymmetric-gate.md): A control point where one direction is cheap and the other expensive — activation energy, ratchets, irreversible commits, CI gates.
- [attractor](https://agentconcepts.io/concepts/attractor.md): A state or set of states in a dynamical system toward which trajectories converge; the basin of attraction is the region of initial conditions that end up there.
- [availability-heuristic](https://agentconcepts.io/concepts/availability-heuristic.md): Estimating the frequency, probability, or importance of a category by the ease with which instances of the category come to mind. Vividness, recency, and personal experience inflate availability; statistical base-rates do not. The shortcut is efficient for genuinely-common categories and badly misca…
- [backpressure](https://agentconcepts.io/concepts/backpressure.md): Flow + feedback: when downstream slows, signal propagates upstream as throttle. Self-regulates the system to the bottleneck rate.
- [batna](https://agentconcepts.io/concepts/batna.md): Best Alternative To a Negotiated Agreement — the walk-away position that constrains what you'll accept. The strength of your BATNA, not your stated position, determines the floor of acceptable outcomes in a negotiation.
- [bike-shedding](https://agentconcepts.io/concepts/bike-shedding.md): Allocating effort inversely to importance — among items competing for a group's or agent's finite attention, the trivial, cognitively-accessible item absorbs the bandwidth the important, hard-to-evaluate item deserves. The mechanism is accessibility, not signal; everyone can opine on the bike-shed's…
- [bookends](https://agentconcepts.io/concepts/bookends.md): Structural symmetry at the boundaries of a unit: the beginning and ending mirror each other, framing what is between and producing closure by recurrence.
- [bootstrapping](https://agentconcepts.io/concepts/bootstrapping.md): A staged self-construction in which the system uses partial outputs of stage N to enable a qualitatively more capable stage N+1 — the seed plus the mechanism by which the system escapes a cold-start problem from inside itself.
- [bottleneck-buffer](https://agentconcepts.io/concepts/bottleneck-buffer.md): Dual along a flow: the bottleneck is the rate-limiting point that determines aggregate throughput; the buffer is the rate-smoothing reservoir that absorbs short-term mismatches.
- [bubble-dynamics](https://agentconcepts.io/concepts/bubble-dynamics.md): Self-reinforcing price (or valuation) growth that decouples from underlying fundamentals via reflexive feedback between expectations and behavior, sustained by new entrants until exhaustion produces a crash. The structural feature is amplifying-feedback with no fundamental anchor and a built-in fail…
- [bulkhead](https://agentconcepts.io/concepts/bulkhead.md): Isolate failure domains so one section's failure cannot sink the whole system; the system is partitioned into independent compartments with no shared resources that could carry failure across boundaries.
- [bundle-vs-isolate](https://agentconcepts.io/concepts/bundle-vs-isolate.md): Design choice of whether to handle addressable work-items in shared execution containers or in separate ones, trading overhead amortization against isolation, per-item specificity, and failure-boundary clarity.
- [caching](https://agentconcepts.io/concepts/caching.md): A close-to-consumer copy of frequently-accessed data; trades staleness for speed. The cache holds a derived view; the authoritative source remains canonical.
- [cadence](https://agentconcepts.io/concepts/cadence.md): The temporal rhythm of a process — the periodic or event-driven beat that governs *when* things happen, distinct from feedback-loop.
- [calibration](https://agentconcepts.io/concepts/calibration.md): An instrument or judgment is aligned against a trusted reference standard so its outputs can be taken at face value; the alignment is a deliberate act that maps raw indications onto true values and quantifies the residual error.
- [call-and-response](https://agentconcepts.io/concepts/call-and-response.md): Dialogic structure where a statement (the call) is followed by an answering statement (the response), with the response shaped by the call and meaningful only in relation to it. The pair — not either alone — is the structural unit; each participant's turn earns its meaning from its placement in the…
- [cargo-cult](https://agentconcepts.io/concepts/cargo-cult.md): Copying the surface shape of a successful thing without the underlying causal force — shape without the load-bearing mechanism.
- [cascade](https://agentconcepts.io/concepts/cascade.md): One event triggers another, which triggers another, in a directed chain where each link's firing is contingent on the previous. The structural shape is initiating-trigger + chain-of-dependent-events + a propagation property that makes each event sufficient to fire the next.
- [catalysis](https://agentconcepts.io/concepts/catalysis.md): Accelerates a reaction or process without being consumed by it; lowers activation energy, increases rate, and emerges unchanged at the end.
- [chain-of-responsibility](https://agentconcepts.io/concepts/chain-of-responsibility.md): A request is passed along a chain of handlers until one of them accepts it; sender does not know which handler will handle it, and handlers do not know who follows them in the chain.
- [chain-of-thought](https://agentconcepts.io/concepts/chain-of-thought.md): Explicit reasoning trace exposed as part of the output, not just consumed internally — "think out loud, then answer." The trace makes the reasoning auditable and often improves the answer itself.
- [channel-capacity](https://agentconcepts.io/concepts/channel-capacity.md): The maximum rate of reliable information transmission through a noisy channel; Shannon's theorem. Capacity is a structural property of the channel (bandwidth, noise), not of the messages or the encoders — and it is a hard ceiling, not a soft target.
- [chekhovs-gun](https://agentconcepts.io/concepts/chekhovs-gun.md): A load-bearing element placed deliberately and early — the placement is notable specifically because the element will be called upon later. "Do not put a loaded rifle on the stage if no one is thinking of firing it."
- [choke-point](https://agentconcepts.io/concepts/choke-point.md): A narrow gateway whose control determines what passes; small physical or logical position with disproportionate strategic value because it constrains flow externally. Externally-controllable, not just rate-limiting.
- [chunking](https://agentconcepts.io/concepts/chunking.md): Active grouping of items into larger units to compress representation for cognitive or computational efficiency; the action that produces a unit-shift in how downstream processes see the content.
- [circuit-breaker](https://agentconcepts.io/concepts/circuit-breaker.md): Auto-cut-off mechanism that stops calling a failing downstream after a threshold of failures is reached; trips OPEN, refuses calls, then probes HALF-OPEN before resuming CLOSED.
- [close-the-loop](https://agentconcepts.io/concepts/close-the-loop.md): A return signal travels back to the specific party that initiated an action, confirming to that originator that the action landed — receipt, status, or acknowledgment routed to the requester, not merely to the system at large. The originator-as-recipient constraint is the concept.
- [closure](https://agentconcepts.io/concepts/closure.md): The perceiver completes incomplete or interrupted patterns by supplying the missing parts. The mind closes gaps — visual, narrative, syntactic, motivational — filling in what is structurally implied but not explicitly present.
- [collider-bias](https://agentconcepts.io/concepts/collider-bias.md): Conditioning on a variable that two others jointly cause induces a spurious association between those two causes, where none existed in the full population. The act of selecting, stratifying, or controlling on the common effect is what creates the association — the structural mirror of a common caus…
- [composite](https://agentconcepts.io/concepts/composite.md): A tree structure where leaves and composites share the same interface and are treated uniformly by clients; recursive containment that hides the depth of the structure.
- [confirmation-bias](https://agentconcepts.io/concepts/confirmation-bias.md): The tendency to seek out, weight, and remember information that confirms existing beliefs while underweighting, ignoring, or actively avoiding information that would disconfirm them. The bias is in the asymmetric handling of evidence relative to the prior, not in the prior itself.
- [confounding](https://agentconcepts.io/concepts/confounding.md): A third variable causes both the apparent cause and the apparent effect, producing an observed association between them that is not a direct causal relationship. The apparent A→B link is real as an association but spurious as a causal claim; the actual causal structure has a common parent driving bo…
- [consensus](https://agentconcepts.io/concepts/consensus.md): Multiple nodes agree on a shared value despite network unreliability and node failures; the agreed-upon value survives even when minority nodes disagree or vanish.
- [conservation-law](https://agentconcepts.io/concepts/conservation-law.md): A quantity that remains invariant under the system's allowed transformations; what goes in must come out or be accounted for.
- [contagion](https://agentconcepts.io/concepts/contagion.md): Failure or behavior propagating across coupled components or institutions via the connections that normally permit beneficial flow; one unit's breakdown spreads to its neighbors and onward through the topology. The structural feature is the coupling that was productive becoming the transmission medi…
- [container](https://agentconcepts.io/concepts/container.md): Image schema: a bounded interior with an inside, an outside, and a boundary that controls what crosses; scope and encapsulation are containment moves.
- [context-asymmetry](https://agentconcepts.io/concepts/context-asymmetry.md): Two contexts that project at different grains or with different visibility; the mismatch is productive when deliberate (information hiding, abstraction boundary) and harmful when accidental (lost signal, missed assumptions).
- [cost-cascade](https://agentconcepts.io/concepts/cost-cascade.md): Cheap-default path handles the common case; expensive-fallback handles the remainder when the cheap path falls short. Cost differential is the value.
- [counterpoint](https://agentconcepts.io/concepts/counterpoint.md): Independent melodic voices in tension AND harmony; each voice is meaningful on its own, and together they form a coherent texture. The concept is the simultaneous independence + coordination — neither alone.
- [dead-reckoning](https://agentconcepts.io/concepts/dead-reckoning.md): Current state is estimated by integrating successive increments of self-motion from a last known fixed position, using no external observation; each increment's error compounds, so total uncertainty grows without bound until an external fix re-anchors the estimate.
- [decorator](https://agentconcepts.io/concepts/decorator.md): A wrapper that adds behavior to a subject without changing its interface; decorators stack to compose behaviors. Behavioral composition without inheritance.
- [defense-in-depth](https://agentconcepts.io/concepts/defense-in-depth.md): Multiple independent defensive layers, each handling some fraction of attacks; the breach of any one isn't catastrophic because the next layer also defends. Stacked-barriers, not isolated-domains.
- [deixis](https://agentconcepts.io/concepts/deixis.md): Reference whose meaning is determined by the context of utterance — "this," "here," "now," "I," "you," "yesterday." The word's denotational content is a pointer; the pointer resolves only against the speaker's situation. Transplant the word without the situation and the reference becomes unresolved…
- [density-dependent-regulation](https://agentconcepts.io/concepts/density-dependent-regulation.md): Effects on a population (or system) that scale with the density of its occupants, producing a regulatory pressure that intensifies as density rises; the higher the density, the stronger the dampening force, until the system stabilizes near a density-dependent carrying capacity. The diagnostic is whe…
- [depth-of-field](https://agentconcepts.io/concepts/depth-of-field.md): Selective focus: a narrow zone is rendered sharp while everything closer or farther recedes into blur. The optical phenomenon names a more general structural move — concentrate clarity on one zone at the cost of clarity elsewhere; the choice of where the sharpness sits IS the directional statement.
- [deus-ex-machina](https://agentconcepts.io/concepts/deus-ex-machina.md): A stuck system suddenly resolved by an external force unprepared by the prior structure; the resolution comes from outside the established rules.
- [differential-diagnosis](https://agentconcepts.io/concepts/differential-diagnosis.md): Systematic narrowing of an explanation-space by enumerating candidate causes and discriminating among them via tests that have different expected results under different candidates. The candidates are known to the practitioner; the search is for which of the known candidates fits the current present…
- [differential-scaling](https://agentconcepts.io/concepts/differential-scaling.md): Coupled quantities grow at different rates as a system scales, so a balance or strategy that holds in the small regime breaks in the large — the divergence of growth exponents (surface ∝ L² but volume ∝ L³) tears the original architecture apart.
- [difficulty-curve](https://agentconcepts.io/concepts/difficulty-curve.md): The progression of demand-on-the-participant over the course of an engagement — how challenge scales as the participant advances. The structural feature is the shape of demand-over-time and its alignment with the participant's growing capability; the curve is well-formed when demand grows just enoug…
- [discovered-attack](https://agentconcepts.io/concepts/discovered-attack.md): A move where one piece stepping aside reveals an attack from a different piece that was already in position behind it. The structural feature is that the moving piece is not the attacker — the attack came from a pre-positioned piece whose threat was occluded by the now-moved piece. The triggering mo…
- [dissonance-and-resolution](https://agentconcepts.io/concepts/dissonance-and-resolution.md): Controlled tension followed by release. A structural element generates a held expectation of unease, instability, or incompleteness; a subsequent element discharges it. The pair — not the tension alone, not the release alone — is the unit. The arc is what gives both ends their meaning.
- [disturbance](https://agentconcepts.io/concepts/disturbance.md): An exogenous event that clears occupied structure and resets a system's developmental trajectory to an earlier state, opening freed substrate for re-colonization while the system retains the capacity to re-traverse the sequence. The reset, not merely the shock, is the concept.
- [doctrine](https://agentconcepts.io/concepts/doctrine.md): An operational rule promoted to a named, addressable artifact: name + triggering condition + protected-against failure + citation address.
- [drift](https://agentconcepts.io/concepts/drift.md): A quantity that is nominally fixed slowly diverges from its reference over time with no restoring force, accumulating unnoticed because nothing announces the change; the divergence is detectable only by re-comparison against the reference.
- [duality](https://agentconcepts.io/concepts/duality.md): Two perspectives on the same underlying object — primal vs. dual, particle vs. wave, system vs. its adjoint — each carrying information the other obscures.
- [eager-vs-lazy](https://agentconcepts.io/concepts/eager-vs-lazy.md): Temporal/scheduling choice for WHEN to do work: eager precomputes and stores; lazy defers until needed. The higher-order concept picks out the cost-time-correctness tradeoff this strategy choice implies.
- [emergence](https://agentconcepts.io/concepts/emergence.md): Collective behavior or properties at one scale arising from simple interactions at a lower scale, not deducible from any single component in isolation.
- [endow](https://agentconcepts.io/concepts/endow.md): Grant an attribute, role, or context to a partner by treating them as having it; the declaration creates the reality. "You're a brain surgeon" makes the partner a brain surgeon for the duration of the scene.
- [entropy](https://agentconcepts.io/concepts/entropy.md): A measure of disorder, uncertainty, or the number of microscopic configurations consistent with a given macroscopic state; tends to increase in closed systems.
- [equilibrium](https://agentconcepts.io/concepts/equilibrium.md): A balance point where opposing forces or rates match, so the system's state stops changing — statically (no movement) or dynamically (flows balance).
- [erosion-deposition](https://agentconcepts.io/concepts/erosion-deposition.md): One transport flow simultaneously removes material from a source site and accumulates it at a sink site; the subtraction and the addition are two ends of a single conserved transfer, so the deposit is diagnostic of the source and the two sites are coupled by the carrier between them.
- [error-correction](https://agentconcepts.io/concepts/error-correction.md): Using redundancy plus structure to detect and recover from corrupted transmission; the *active reconstruction* of intended signal from noisy data. Distinct from error-detection — correction reconstructs the original, not just flags the deviation.
- [evaluator-optimizer](https://agentconcepts.io/concepts/evaluator-optimizer.md): Generator produces output, evaluator critiques it, generator revises — an iterated loop that compounds quality across rounds. The diagnostic: "is the critique getting acted on, or just produced?"
- [event-sourcing](https://agentconcepts.io/concepts/event-sourcing.md): Derive current state by replaying a log of events; the log is the source of truth, current state is a materialized view. The history IS the data.
- [exaptation](https://agentconcepts.io/concepts/exaptation.md): A feature that evolved (or was built) for one purpose is co-opted for another, often unintended use; the existing structure is repurposed rather than redesigned. Unplanned-reuse is the distinguishing character.
- [externalities](https://agentconcepts.io/concepts/externalities.md): Costs or benefits that fall on parties outside a transaction and are not captured in the transaction's price; the consequential effects of an act flow to non-participants without their consent or compensation. Negative and positive externalities are paired contrasts at the polarity axis.
- [eye-line-match](https://agentconcepts.io/concepts/eye-line-match.md): A signal (a character looking at something off-screen) is followed by the referent (a cut to what they're looking at), with the audience supplying the inference that joins them. The structure exports beyond cinema: any signal-then-referent pair where the audience completes the bridging inference ope…
- [facade](https://agentconcepts.io/concepts/facade.md): A simplified front-end that hides the complexity of a subsystem behind a single coherent interface; consumers interact with the facade and never see the machinery behind it.
- [feedback-loop](https://agentconcepts.io/concepts/feedback-loop.md): Output influences future input through a returning signal; the loop self-reinforces (positive polarity) or self-regulates (negative polarity).
- [few-shot](https://agentconcepts.io/concepts/few-shot.md): A small set of examples teaches the pattern; the agent generalizes from demonstration rather than from explicit description. "Watch this and that, then do the next one."
- [figure-ground](https://agentconcepts.io/concepts/figure-ground.md): Perceptual organization separating an attended foreground (the figure) from a receding context (the ground). The split is constructed by the observer, not given by the stimulus — which means the same input can be re-segmented, and what counts as figure depends on attention, framing, and convention.
- [find-the-game](https://agentconcepts.io/concepts/find-the-game.md): Identify the schema-anomaly in a situation, treat it as the load-bearing defining feature, then project it analogically to other moves or domains. The improv-canonical move that generalizes far beyond improv — it's how science finds paradigms, debuggers find bugs, doctors diagnose, and artists make…
- [fixed-point](https://agentconcepts.io/concepts/fixed-point.md): A point where input equals output under an operation: f(x) = x. The system stops moving here; the operation maps it to itself.
- [flow](https://agentconcepts.io/concepts/flow.md): Directed movement of items, work, energy, or signal from a source through a medium to a destination; the unit of throughput.
- [force-multiplier](https://agentconcepts.io/concepts/force-multiplier.md): A small input amplified into a large output via leverage, automation, or composition; the structural shape that makes asymmetric impact possible. Amplification by output magnitude, not by acceleration.
- [foreshadowing](https://agentconcepts.io/concepts/foreshadowing.md): Early signal that something is coming — primes the observer's expectation without yet delivering. The signal's shape constrains what arrival will count.
- [fork](https://agentconcepts.io/concepts/fork.md): A single move that creates multiple simultaneous threats or effects, forcing the defender to choose which to address; the defender can only respond to one per cycle, so loss on at least one front is structurally guaranteed.
- [frame-story](https://agentconcepts.io/concepts/frame-story.md): Nested narrative-as-container: an outer story contains an inner story (which may itself contain another). The frame both bounds and contextualizes what it contains; the inner content has different rules than the outer.
- [framing-effect](https://agentconcepts.io/concepts/framing-effect.md): Equivalent options are chosen differently depending on how they are presented. The same outcome, framed as a gain versus a loss, or as a probability of success versus a probability of failure, systematically produces different choices despite the underlying facts being identical. The empirical pheno…
- [gate](https://agentconcepts.io/concepts/gate.md): A controlled passage point that permits or blocks what flows through it according to a condition evaluated by a controller.
- [graceful-degradation](https://agentconcepts.io/concepts/graceful-degradation.md): Partial-failure-acceptable mode: when a component or dependency fails, the system continues with reduced functionality rather than total failure. The fallback path is designed-in, not improvised.
- [gradient](https://agentconcepts.io/concepts/gradient.md): Anti-symmetry schema: a dimension with direction — cheap one way, expensive the other; one state attracts, another repels.
- [graduation-promotion](https://agentconcepts.io/concepts/graduation-promotion.md): Scaffolding to adult form. A provisional artifact becomes the canonical version at a qualitative threshold, not via gradual drift.
- [grain](https://agentconcepts.io/concepts/grain.md): The level of resolution at which an entity is individuated — its identity-grain. Confused grains produce locally-correct but aggregation-wrong bugs.
- [group-mind](https://agentconcepts.io/concepts/group-mind.md): Emergent collective coordination from an ensemble of individuals each operating with limited local information; the group becomes smarter than the sum of its members.
- [half-renamed-codebase](https://agentconcepts.io/concepts/half-renamed-codebase.md): Terminology lag as project-maturation signal; the lagging name surfaces an unabsorbed mental model that the project has structurally outgrown.
- [heightening](https://agentconcepts.io/concepts/heightening.md): Raising the stakes of an established pattern via analogical extension within consistent constraints. Operates within-scene (escalating an in-situ pattern) OR across-domain (escalating a pattern projected via find-the-game). Independent of how the pattern got established; heightening is the stake-esc…
- [hindsight-bias](https://agentconcepts.io/concepts/hindsight-bias.md): Past events appear more predictable in retrospect than they were in foresight. After learning the outcome, the observer reconstructs their prior estimate to better match what actually happened — "I knew it all along" — and uses the reconstructed estimate to evaluate the original decision. The recons…
- [hoist-by-own-petard](https://agentconcepts.io/concepts/hoist-by-own-petard.md): An agent harmed by their own deliberate construction or device — the bomb they made blows them up; the trap they set catches them; the doctrine they advocated for backfires on them.
- [hysteresis](https://agentconcepts.io/concepts/hysteresis.md): Output depends not just on current input but on the history of inputs — the path matters, not just the current state.
- [idempotency](https://agentconcepts.io/concepts/idempotency.md): Same operation produces the same outcome regardless of how many times it is applied. f(f(x)) = f(x); the second application is a no-op.
- [in-medias-res](https://agentconcepts.io/concepts/in-medias-res.md): Entry into a process at the middle, not the beginning; backfill, context, or origin arrives after the audience/consumer is already in motion. The entry-point is mid-cycle, not pre-cycle.
- [Concepts](https://agentconcepts.io/concepts/index.md): Browse the full catalog of structural primitives.
- [inertia](https://agentconcepts.io/concepts/inertia.md): Resistance to change in motion or state; objects keep doing what they are already doing unless acted on by a force proportional to the change desired.
- [internalization](https://agentconcepts.io/concepts/internalization.md): External structure's form is taken up as the system's own internal structure, so the behavior the external support produced is now self-generated; the form crosses the boundary from outside the system to inside it.
- [isomorphism](https://agentconcepts.io/concepts/isomorphism.md): Same structure, different content; a bijective structure-preserving map between two systems such that operations in one have exact counterparts in the other.
- [kaizen](https://agentconcepts.io/concepts/kaizen.md): Continuous small improvements at every level; many small wins compounding over time rather than infrequent large overhauls. The discipline is the small-step cadence, not the size of any single improvement.
- [kanban](https://agentconcepts.io/concepts/kanban.md): Visualize work-in-progress and signal demand upstream; the production line pulls only what downstream actually needs. Make the work visible, cap how much is in flight, let downstream pace the upstream.
- [kayfabe](https://agentconcepts.io/concepts/kayfabe.md): Maintained fiction that all parties know is fiction but agree to treat as real for the duration of the engagement. Distinct from deception (which requires asymmetric belief) — kayfabe requires symmetric knowledge plus a shared protocol to act-as-if.
- [kernel](https://agentconcepts.io/concepts/kernel.md): A small, essential, generative core from which a system's outer behavior unfolds — the part everything else runs on or grows out of, distinct from the husk around it.
- [keystone-species](https://agentconcepts.io/concepts/keystone-species.md): A component that is small in population or size but disproportionate in structural impact — remove it and the system collapses. Load-bearing + small-relative-to-impact is the distinguishing combination.
- [kingmaker-problem](https://agentconcepts.io/concepts/kingmaker-problem.md): A situation in which an actor who cannot win the contest themselves nonetheless has decisive influence over which of the remaining contenders does win. The structural feature is the asymmetry between two capabilities: the kingmaker's capability-to-influence-outcome is high, but their capability-to-w…
- [leaky-abstraction](https://agentconcepts.io/concepts/leaky-abstraction.md): Container + projection: every abstraction that hides a substrate eventually leaks because projections are non-invertible. The "leak" is the substrate bleeding through the boundary.
- [learning-curve](https://agentconcepts.io/concepts/learning-curve.md): The progression of capability acquired over engagement with a domain — how skill scales with practice, time, or repetition. The structural feature is the shape of capability-over-time, including its typical non-linearities: slow start (the basics threshold), rapid improvement (the productive middle)…
- [lifecycle-cost](https://agentconcepts.io/concepts/lifecycle-cost.md): The true cost of a thing is its full cradle-to-grave total — acquisition plus operation, maintenance, and disposal — most of which is hidden at the point of purchase, so the visible sticker price systematically understates what the thing actually costs over its life.
- [liminality](https://agentconcepts.io/concepts/liminality.md): The transitional state of being between defined categories, roles, or phases — neither one nor the other, governed by different rules than either bracketing state. Coined by Arnold van Gennep, developed by Victor Turner: the threshold-position where ordinary structure is suspended and the participan…
- [load-balancing](https://agentconcepts.io/concepts/load-balancing.md): Distribute work across multiple workers by some routing function; the worker pool absorbs aggregate load that no single worker could carry, with the router as the always-traversed dispatch point.
- [load-bearing](https://agentconcepts.io/concepts/load-bearing.md): Element actually carrying weight in a system; "what if I removed this?" diagnostic.
- [local-minimum](https://agentconcepts.io/concepts/local-minimum.md): A point in an optimization landscape where every nearby direction looks worse, but the global landscape has better attractors elsewhere. The trap of locally-optimal stuck.
- [loop-completion](https://agentconcepts.io/concepts/loop-completion.md): Gaps become visible precisely because the journey is coherent enough to reveal them. Pre-coherent vs post-coherent regime diagnostic.
- [loss-aversion](https://agentconcepts.io/concepts/loss-aversion.md): Losses loom larger than equivalent gains in subjective valuation. The same magnitude of outcome (e.g., \$100) carries roughly 2× the psychological weight when framed as a loss as when framed as a gain. The asymmetry is structural and load-bearing for prospect theory.
- [macguffin](https://agentconcepts.io/concepts/macguffin.md): A driver of plot or action whose specific content is arbitrary — what matters is that the characters/agents want it, not what it actually is. The narrative engine that runs equally well on any content-filler.
- [make-wrong-unrepresentable](https://agentconcepts.io/concepts/make-wrong-unrepresentable.md): Replace runtime gate with structural absence: the invalid state has no handler, no representation, no construct — not merely blocked.
- [manifold](https://agentconcepts.io/concepts/manifold.md): Simple in any small region, but not as a whole: every neighborhood is faithfully captured by a simple local model, yet no single model covers the whole — describing all of it forces a choice between one distorted global model or many local models stitched together. Rules that hold locally fail at gl…
- [marginal-vs-average](https://agentconcepts.io/concepts/marginal-vs-average.md): The diagnostic distinction between the value of the next unit (marginal) and the value across all units to date (average); rational decisions about additions, removals, and continuations require the marginal, not the average. Mistaking one for the other produces systematic decision errors.
- [mean-reversion](https://agentconcepts.io/concepts/mean-reversion.md): Deviation from a baseline generates a restoring pressure back toward the baseline; the further the deviation, the stronger the pulling-back force. The dynamic — not the endpoint — is the concept.
- [mediator](https://agentconcepts.io/concepts/mediator.md): A central hub that coordinates interactions among a set of peers; peers communicate through the mediator rather than directly, reducing the N²-edge tangle to N spokes.
- [mise-en-place](https://agentconcepts.io/concepts/mise-en-place.md): Everything in its place; pre-arranged preparation that pays in downstream execution speed. The cost is paid before the pressure moment; the payoff is sustained tempo during the pressure moment.
- [mixed-use](https://agentconcepts.io/concepts/mixed-use.md): Multiple functions sharing a single zone — residential plus commercial plus recreational in one neighborhood; diversity-of-function in one container. The mixing produces emergent properties (street life, redundancy, walkability) that single-use cannot.
- [modulation](https://agentconcepts.io/concepts/modulation.md): Change of key as structural shift; a deliberate move to a related-but-different frame. The new key is reachable from the old via shared pivots, but once arrived in, the listener / observer operates by the new frame's rules.
- [momentum](https://agentconcepts.io/concepts/momentum.md): Trend persistence — a system's recent direction of motion is itself evidence that motion will continue in the same direction. Past returns predict future returns of the same sign within a regime, and the predictive strength rises with directional consistency. The amplifying-trend specialization of i…
- [moral-hazard](https://agentconcepts.io/concepts/moral-hazard.md): An actor insulated from the cost of their risk-taking takes more risk than they would if they bore the full downside; the insulation is what produces the riskier behavior.
- [motif](https://agentconcepts.io/concepts/motif.md): Recurring small pattern that carries meaning across instances; not a single setup-and-payoff, but a self-similar element that compounds significance through repetition. Each occurrence accumulates the meaning of the previous ones.
- [multi-channel-ingest](https://agentconcepts.io/concepts/multi-channel-ingest.md): Multiple input channels with distinct trust / latency / volume / cost profiles converge into a unified store with reconciliation logic.
- [multi-hop-routing](https://agentconcepts.io/concepts/multi-hop-routing.md): Items transit intermediate hubs from source to destination; each hop knows only the next hop, not the final destination — the routing topology is distinct from the cargo and from any single hop.
- [mutualism](https://agentconcepts.io/concepts/mutualism.md): Symbiotic mutual benefit between two parties — each contributes something the other needs; both gain. Same multi-actor coordination structure as tragedy-of-commons, but with payoff polarity reversed.
- [negative-space](https://agentconcepts.io/concepts/negative-space.md): The area, time, or position not occupied is structurally load-bearing — the absence carries meaning. Figure-ground inversion: what is missing defines what is there.
- [network-effect](https://agentconcepts.io/concepts/network-effect.md): Value grows with number of participants; the Nth participant adds value to all N-1 previous participants. Positive-feedback at the participation level.
- [niche](https://agentconcepts.io/concepts/niche.md): A specific role within a larger system — a particular intersection of conditions where one species, actor, or product specializes. Distinct from grain (level of resolution): niche is the role-within-system, grain is the resolution-level.
- [niche-partitioning](https://agentconcepts.io/concepts/niche-partitioning.md): The multi-agent dynamic by which similar would-be competitors avoid competitive exclusion by specializing into distinct sub-regions of resource space; coexistence emerges not from cooperation but from each occupant's adaptation to a non-overlapping slice. Distinct from niche (the role) by being the…
- [observer](https://agentconcepts.io/concepts/observer.md): A subject maintains a list of dependents (observers) and notifies them automatically when its state changes; pub-sub at the object level. Decouples what-changed from who-cares.
- [one-way-ratchet](https://agentconcepts.io/concepts/one-way-ratchet.md): A doctrine that guarantees correctness via monotonic growth (always-add, never-remove) — and therefore requires a counter-doctrine for periodic prune-down to prevent unbounded accumulation.
- [opportunity-cost](https://agentconcepts.io/concepts/opportunity-cost.md): The value of the next-best foregone alternative when a choice is made; the cost of a decision is not just what you pay but what you would have gotten otherwise. The diagnostic shifts attention from explicit cost to the displaced alternative the choice rules out.
- [orchestrator-workers](https://agentconcepts.io/concepts/orchestrator-workers.md): A central orchestrator decomposes a complex task into sub-tasks and delegates each to specialized workers; the orchestrator aggregates the results. Decomposition + delegation + aggregation is the trifecta.
- [parallel-vs-serial](https://agentconcepts.io/concepts/parallel-vs-serial.md): The scheduling choice of whether operations run concurrently or sequentially; trades latency reduction against coordination overhead, ordering guarantees, and resource contention.
- [pareto-principle](https://agentconcepts.io/concepts/pareto-principle.md): A small minority of inputs accounts for a large majority of outputs; the vital few dominate the trivial many. The load-bearing feature is concentration of contribution across a population of contributors, not mere non-linearity or diminishing returns.
- [partial-order](https://agentconcepts.io/concepts/partial-order.md): An order relation where some pairs are genuinely incomparable, so the structure is a branching web rather than a line — and forcing it onto a line destroys information.
- [phase-transition](https://agentconcepts.io/concepts/phase-transition.md): A qualitative, often discontinuous change in system behavior at a critical threshold of a control parameter; below the threshold one regime, above it another.
- [pin](https://agentconcepts.io/concepts/pin.md): A visible threat that constrains a piece's mobility because something more valuable is positioned behind it. The pinned piece is not physically blocked — it could move — but moving exposes the more-valuable piece to capture or damage, making the move effectively forbidden.
- [poka-yoke](https://agentconcepts.io/concepts/poka-yoke.md): Mistake-proofing; designing the physical or procedural system so the wrong action is structurally impossible or immediately visible. The defect cannot occur, or if it does, it cannot pass unnoticed.
- [polyrhythm](https://agentconcepts.io/concepts/polyrhythm.md): Multiple rhythms operating simultaneously, each with its own period, none reducible to the others. The structure is not just multi-voice (that's counterpoint) and not just multi-beat (that's a complex single meter) — it is genuinely multi-period, with the rhythms periodically coinciding and periodic…
- [polysemy](https://agentconcepts.io/concepts/polysemy.md): One word, name, or symbol with multiple *related* senses — a single form that has been extended by analogy, metaphor, or specialization across uses. Distinct from homonymy (multiple unrelated senses sharing a form by accident); polysemy's senses share a structural family resemblance that makes the m…
- [pragmatic-implicature](https://agentconcepts.io/concepts/pragmatic-implicature.md): Meaning conveyed beyond what is literally said, derived by the listener via assumptions of cooperative speech. The literal content underspecifies the message; the listener infers the rest from context + the assumption that the speaker is being relevant, informative, truthful, and clear (Grice's maxi…
- [pre-coherent-vs-post-coherent-regime](https://agentconcepts.io/concepts/pre-coherent-vs-post-coherent-regime.md): A regime classifier: the shape of feedback the system is producing tells you which regime you're in — pre-coherent (architecture problems surface first) or post-coherent (loop-completion gaps surface). Different regimes warrant different responses.
- [precedent](https://agentconcepts.io/concepts/precedent.md): A prior decision binds or guides a current decision via structural similarity; case-based reasoning made institutional. The current case is decided by analogy to a previously-decided case held to be relevantly similar.
- [predator-prey-dynamics](https://agentconcepts.io/concepts/predator-prey-dynamics.md): Coupled oscillation between two populations linked by an asymmetric one-eats-the-other relationship; prey-growth drives predator-growth, predator-growth drives prey-decline, prey-decline drives predator-decline, prey-recovery restarts the cycle. The structural feature is bidirectional coupling with…
- [prestige-loop](https://agentconcepts.io/concepts/prestige-loop.md): A meta-progression pattern in which within-cycle progress is periodically reset to start over, while a separate cross-cycle reward layer accumulates permanently. Each cycle begins from a near-baseline but with permanent benefits earned from previous cycles. The structural feature is two-layer progre…
- [principal-agent](https://agentconcepts.io/concepts/principal-agent.md): A delegator (principal) hires a delegate (agent) to act on their behalf; the agent has potentially misaligned incentives and acts in their own interest, sometimes against the principal's. Misalignment + information-asymmetry is the core structural shape.
- [prisoners-dilemma](https://agentconcepts.io/concepts/prisoners-dilemma.md): A cooperate-vs-defect choice where defecting is individually rational but mutual cooperation is collectively better; the game's payoff structure produces the bad equilibrium even when both actors prefer cooperation.
- [prodrome](https://agentconcepts.io/concepts/prodrome.md): Early, non-specific signs or symptoms that appear before a condition's full, characteristic presentation; the "something is coming" stage where the signal is real but the pattern is not yet diagnostic. The prodrome is recognized backward more reliably than forward — most prodromes are only retrospec…
- [prompt-chaining](https://agentconcepts.io/concepts/prompt-chaining.md): Sequential composition: each stage's output is the next stage's input. The pipeline does as a whole what no single stage can do alone.
- [prophylaxis](https://agentconcepts.io/concepts/prophylaxis.md): Preventing the opponent's plan before executing your own; defensive moves that constrain the future state-space rather than gaining immediate advantage. The cost is paid now to remove options the adversary would otherwise have.
- [proposal-judgment-split](https://agentconcepts.io/concepts/proposal-judgment-split.md): Generating candidates and judging them are assigned to separate roles rather than fused in one, because the two roles have opposed optima — the proposer optimizes coverage, the judge optimizes precision — and fusing them collapses the discrimination the judge is supposed to supply.
- [proximity](https://agentconcepts.io/concepts/proximity.md): Items near each other in space, time, or some other dimension tend to be perceived (and treated) as belonging together. Distance is a grouping signal; nearness implies relatedness even in the absence of explicit relational evidence.
- [proxy](https://agentconcepts.io/concepts/proxy.md): A stand-in that controls access to a real subject while preserving the subject's interface; the client thinks they're talking to the subject directly, but the proxy intercepts.
- [queue](https://agentconcepts.io/concepts/queue.md): FIFO buffer for asynchronous work; decouples producer from consumer in time, rate, and identity. The queue absorbs producer bursts and feeds consumers at their own pace.
- [quietly-load-bearing](https://agentconcepts.io/concepts/quietly-load-bearing.md): Failure-mode-load-bearing distinct from decision-time-load-bearing; the load-bearing diagnostic doesn't run cleanly when failure surfaces only at delayed retrospective.
- [quorum](https://agentconcepts.io/concepts/quorum.md): The minimum subset of participants whose joint agreement constitutes a legitimate decision; below the threshold the system cannot bind, above it the decision stands even over the dissent of those absent or opposed.
- [rabbit-hole](https://agentconcepts.io/concepts/rabbit-hole.md): A tangential descent that deepens past its value, carrying the pursuer away from the original objective. Each level reveals more to pursue; the original question goes unanswered at the bottom.
- [race-condition](https://agentconcepts.io/concepts/race-condition.md): Two paths + shared container: two parallel paths converge on a contended resource without coordination, producing outcomes that depend on timing rather than logic.
- [rate-limiting](https://agentconcepts.io/concepts/rate-limiting.md): Bounded rate of consumption or production: "you may not do more than N of X per Y." A hard cap on flow, enforced by the system that owns the resource.
- [reciprocity](https://agentconcepts.io/concepts/reciprocity.md): Mutual exchange that creates a normative obligation — receipt of something triggers an expectation, often unspoken, that the recipient give back in turn. Distinct from mutualism (which is structurally mutual without depending on norms) by the obligation-creation mechanism.
- [recursion](https://agentconcepts.io/concepts/recursion.md): A structure or process is defined in terms of smaller instances of itself, with a base case that terminates the regress; the whole is generated by repeatedly reducing to a simpler copy of the same form until the base case is reached.
- [red-herring](https://agentconcepts.io/concepts/red-herring.md): A load-bearing distractor: an element placed (or surfacing organically) that the observer takes to be load-bearing, but is not. Structural sibling of chekhov's-gun — looks the same on the surface; the difference is the payoff.
- [redundancy](https://agentconcepts.io/concepts/redundancy.md): Repeating information so noise-induced loss is recoverable; extra structure that costs space (bits, money, time) but buys robustness. The diagnostic: the redundancy is load-bearing precisely when the noise actually fires.
- [reference-vs-value](https://agentconcepts.io/concepts/reference-vs-value.md): The choice between holding a thing by pointing at the canonical original (reference: shared identity, changes propagate) versus holding an independent copy (value: isolated, changes don't propagate). The tradeoff between two modes of holding, not either mode alone.
- [reflection](https://agentconcepts.io/concepts/reflection.md): An agent evaluates its own output and revises based on that self-critique; the generator and evaluator collapse into one actor playing two roles.
- [reframe](https://agentconcepts.io/concepts/reframe.md): Deliberate change of the perceptual or conceptual context applied to a situation, such that the interpretation changes without the underlying facts changing. The operation is on the schema, not on the events the schema is being applied to.
- [register-shift](https://agentconcepts.io/concepts/register-shift.md): Deliberate or context-driven switch in style, formality, or sub-language across the same speaker / system / interface — from technical to lay, formal to casual, clinical to bedside, internal to public-facing. The semantic content can be nearly preserved while the surface style transforms; what chang…
- [replication](https://agentconcepts.io/concepts/replication.md): Copy the same data across multiple stores for redundancy and read-throughput. Each replica holds the full dataset; reads can be served from any; writes propagate to all.
- [resonance](https://agentconcepts.io/concepts/resonance.md): Amplification when a driving force matches the natural frequency of an oscillating system; small periodic inputs produce large responses at the right frequency.
- [retry-with-backoff](https://agentconcepts.io/concepts/retry-with-backoff.md): Repeat a failed operation with increasing delays between attempts; prevents retry-storms that would otherwise compound transient failures into systemic ones.
- [reverse-causation](https://agentconcepts.io/concepts/reverse-causation.md): The causal arrow runs opposite to the assumed direction — B causes A, not A causes B. The observed association is real and needs no third variable; the error is purely in the imputed direction of causation.
- [ritual-phases](https://agentconcepts.io/concepts/ritual-phases.md): Three-phase structure that organizes consequential transitions — separation, transition (liminality), incorporation. Coined by Arnold van Gennep for rites of passage; the structural pattern recurs across non-ritual contexts wherever a participant must move between defined states with social meaning.…
- [rivals-into-router](https://agentconcepts.io/concepts/rivals-into-router.md): "A vs B" rival investments reframe as a router: dispatch to A or B conditioned on situation-shape, rather than choosing one.
- [root-cause-analysis](https://agentconcepts.io/concepts/root-cause-analysis.md): Tracing a symptom or failure backward through its causal chain to the deepest fixable origin — the node that, once addressed, collapses the whole downstream tree of effects. Fix the cause, not the symptom; the countermeasure goes at the root, not at the surface.
- [route-as-context](https://agentconcepts.io/concepts/route-as-context.md): The path encodes meaning. When a mode flag or query param fights the route, the route is wrong — route identity should carry intent.
- [routing](https://agentconcepts.io/concepts/routing.md): An incoming request is classified, then dispatched to one of several specialized handlers that share a common output contract. The structural shape is classifier + handler-set + dispatch — choose-which-once, then forward; the caller is transparent to which handler ran.
- [saga](https://agentconcepts.io/concepts/saga.md): Long-running transaction broken into a sequence of local sub-steps with compensating actions; if a sub-step fails, run compensations in reverse to undo the partially-applied work.
- [satisficing](https://agentconcepts.io/concepts/satisficing.md): Strategic acceptance of a good-enough option rather than continued search for the optimum; the bounded-rationality response when full optimization is infeasible or its cost exceeds the marginal gain.
- [saturation](https://agentconcepts.io/concepts/saturation.md): Diminishing returns approaching an upper bound; each additional unit of input produces less additional output as the system approaches its capacity.
- [scaffolding](https://agentconcepts.io/concepts/scaffolding.md): External, transient support that holds a target system in place while it becomes self-supporting, then is removed — distinguished from permanent dependencies by the explicit removal criterion and from self-construction by being external to the target system.
- [schema-anomaly](https://agentconcepts.io/concepts/schema-anomaly.md): Recognition of where something in a system does not fit the expected schema for the situation; the noticed deviation is itself the load-bearing thing that makes this instance distinct from the generic case.
- [seam](https://agentconcepts.io/concepts/seam.md): Boundary between two systems with different native formats, conventions, or assumptions; where translation happens and where most bugs live.
- [seeding](https://agentconcepts.io/concepts/seeding.md): A small initial input that disproportionately determines the emergent shape of what grows from it; high-leverage choice whose effect persists past its scale.
- [selection-bias](https://agentconcepts.io/concepts/selection-bias.md): The sample being analyzed does not represent the population the conclusion is meant to apply to, because the selection mechanism is associated with the variables being studied. The data is real but the inferences from it generalize wrongly. Includes survivorship bias, response bias, self-selection,…
- [serialization](https://agentconcepts.io/concepts/serialization.md): A reversible transform between an active, executable form and an inert, portable form that can be stored or transmitted and later reconstituted back into the active form.
- [shape](https://agentconcepts.io/concepts/shape.md): Abstract structural form of a problem, independent of surface content. The recognition mechanism that lets cross-domain analogies feel right.
- [sharding](https://agentconcepts.io/concepts/sharding.md): Partition a dataset across multiple stores by a key function; "split then route." Each shard holds a disjoint slice; reads/writes are routed to the owning shard via the key.
- [similarity](https://agentconcepts.io/concepts/similarity.md): Items that share surface features — color, shape, voice, style, tag — tend to be perceived (and treated) as belonging together. Surface resemblance is itself a grouping signal, distinct from and prior to any claim of structural or relational sameness.
- [simpsons-paradox](https://agentconcepts.io/concepts/simpsons-paradox.md): An aggregate trend across a population reverses direction when the population is stratified by a third variable. The same numeric data supports one conclusion at the aggregate level and the opposite conclusion at the stratum level. Which conclusion is correct depends on the causal structure, not on…
- [skewer](https://agentconcepts.io/concepts/skewer.md): A tactical attack along a line where a more-valuable piece sits in front and a less-valuable piece sits behind. The attacker threatens the more-valuable piece, which must move to safety, exposing the less-valuable piece behind it to capture. The structural feature is forced-disclosure-via-displaceme…
- [snowball-effect](https://agentconcepts.io/concepts/snowball-effect.md): A dynamic in which advantage held at one time causally produces additional advantage at later times; the advantage compounds from itself rather than from external addition. The structural feature is that gain at T enables more gain at T+1 — the input to growth is the previously-accumulated output, n…
- [spike](https://agentconcepts.io/concepts/spike.md): A time-boxed investigative task whose purpose is to reduce uncertainty before committing to a build; output is understanding, not shipped product.
- [stack-layer](https://agentconcepts.io/concepts/stack-layer.md): Abstraction levels built on each other, each depending on the correctness of the layer below. Which layer is this problem really at?
- [strategy](https://agentconcepts.io/concepts/strategy.md): A family of interchangeable algorithms encapsulated behind a common interface, selected at runtime; the context delegates to a strategy object rather than coding the algorithm inline.
- [substitution](https://agentconcepts.io/concepts/substitution.md): One thing standing in for another in a role — different filler, same need; the new occupant performs the function the original occupant performed, often with different mechanism but equivalent effect at the consuming interface. Distinct from exaptation (which is same filler, different function).
- [substrate-surface-amplifier](https://agentconcepts.io/concepts/substrate-surface-amplifier.md): Three-element layered architecture with named roles + build-order: substrate (moat) → surface (delivery) → amplifier (compounds over time).
- [succession](https://agentconcepts.io/concepts/succession.md): Sequential development of a system through structured stages, each transforming the substrate in ways that enable (and require) the next stage; the path from disturbance to mature community runs through a series of qualitatively distinct intermediate regimes, not a continuous gradient.
- [sunk-cost-fallacy](https://agentconcepts.io/concepts/sunk-cost-fallacy.md): Letting unrecoverable prior investment influence a forward decision whose payoff depends only on future costs and benefits. The fallacy is the leakage of past expenditure into a forward-looking calculation it should not enter; the more was spent, the more tempting the leak.
- [surface](https://agentconcepts.io/concepts/surface.md): What's exposed to the outside — image-schema sense (UI, top layer) and force-dynamic sense (attack, friction) mediating internal/external.
- [sustainable-yield](https://agentconcepts.io/concepts/sustainable-yield.md): A renewable stock regenerates at some rate; extraction is sustainable only while the take stays at or below the regeneration rate. Exceeding it draws down the stock itself, which lowers future regeneration, so over-extraction is self-accelerating toward collapse.
- [symmetry](https://agentconcepts.io/concepts/symmetry.md): Invariance of a structure under a specified transformation; what stays the same when you do something to it.
- [syncopation](https://agentconcepts.io/concepts/syncopation.md): Accents falling on weak beats or off-beats — the deliberate displacement of emphasis from where the underlying meter expects it. The effect requires the meter to be established first; the syncopation derives its force from the violation of an expectation the listener has been taught to hold.
- [syndromic-presentation](https://agentconcepts.io/concepts/syndromic-presentation.md): A cluster of co-occurring signs, symptoms, or observations that together form a recognizable pattern naming the underlying condition, even when no single element is diagnostic on its own. The pattern is load-bearing because the conjunction is more informative than any element alone; pattern-recognit…
- [taboo](https://agentconcepts.io/concepts/taboo.md): Strong cultural prohibition whose violation triggers ritualized response (revulsion, shaming, expulsion, pollution-restoration ceremonies). Distinct from ordinary rules by the affect-and-response polarity: rules carry consequences; taboos carry charged consequences that mobilize collective response.…
- [template-method](https://agentconcepts.io/concepts/template-method.md): An algorithm skeleton with overridable steps: the outer structure is fixed, and specific steps (the "hook" points) can be customized by specialized implementations.
- [tempo](https://agentconcepts.io/concepts/tempo.md): Initiative as a strategic resource — whose move it is matters as much as what is on the board. A move can gain or lose tempo independently of its effect on material; the right to act next is itself a value.
- [tipping-point](https://agentconcepts.io/concepts/tipping-point.md): Threshold crossing where a small parameter change produces a discontinuous AND irreversible state change. The irreversibility constraint is what distinguishes a tipping-point from a generic phase-transition.
- [total-order](https://agentconcepts.io/concepts/total-order.md): Every pair of elements is comparable under one relation, so the elements fall onto a single line — N items resolved into one unique sequence.
- [trade-off](https://agentconcepts.io/concepts/trade-off.md): A choice between two or more desirable outcomes that cannot be simultaneously maximized; gain along one dimension requires loss along another. The structural shape is a frontier of best-simultaneous-achievements (Pareto) plus a forward choice point at which the agent commits to one position on the f…
- [tragedy-of-commons](https://agentconcepts.io/concepts/tragedy-of-commons.md): Individually-rational decisions yielding collectively-destructive outcomes when a resource is shared without coordination; each actor's private incentive points opposite the collective good.
- [transitivity](https://agentconcepts.io/concepts/transitivity.md): A relation's chaining-inference license: when R(A,B) and R(B,C) hold, R(A,C) follows. The concept is the license and its failure modes, not any one relation.
- [trigger-rule-pair](https://agentconcepts.io/concepts/trigger-rule-pair.md): A heuristic earns its keep only when paired with an invocation condition: 'When [situation S], apply [rule R]' makes the rule addressable.
- [uniformity-dividend](https://agentconcepts.io/concepts/uniformity-dividend.md): Conserving shape across N instances keeps cost roughly constant rather than linear in N. The dividend compounds with cardinality.
- [vertical-slice](https://agentconcepts.io/concepts/vertical-slice.md): Truncate one dimension, preserve another. Build a narrow vertical column (full depth, narrow breadth) instead of a complete horizontal layer first.
- [walk-away-point](https://agentconcepts.io/concepts/walk-away-point.md): The worse-than-which I refuse; an explicit threshold below which I exit rather than continue. Pre-committed in advance of the moment of choice, so the decision under pressure is mechanical rather than negotiated.
- [wisdom-of-crowds](https://agentconcepts.io/concepts/wisdom-of-crowds.md): Aggregation of independent estimates produces accuracy or quality exceeding any individual estimate; the independence is constitutive — correlated estimates break the effect.
- [write-ahead-log](https://agentconcepts.io/concepts/write-ahead-log.md): Record intent-of-change to a durable log before applying it; the log becomes the source of truth for atomicity, recovery, and replay.
- [yes-and](https://agentconcepts.io/concepts/yes-and.md): Constructive collaborative move: accept the partner's offer (yes) and extend it (and). The acceptance is non-negotiable; the extension is what compounds value across turns.
- [zero-sum](https://agentconcepts.io/concepts/zero-sum.md): A fixed total is divided among parties, so one party's gain is exactly another's loss; the sum of all payoffs is constant regardless of how the parties act. There is no joint move that makes everyone better off — only redistribution.
- [zugzwang](https://agentconcepts.io/concepts/zugzwang.md): Forced to move when any move worsens position; the *obligation to act* IS the disadvantage. Distinct from "stuck in a bad place" — zugzwang is the structural compulsion to leave a tolerable state for a worse one because passing is not allowed.
- [Agent Concepts Overview](https://agentconcepts.io/overview/index.md): A structural-reasoning counterpart to [Agent Skills](https://agentskills.io).

## OpenAPI Specs

- [openapi](https://agentconcepts.io/api-reference/openapi.json)