What Is a Context Gap?
A context gap is the difference between the state of the world an automated decision acts on and the actual state of the world at the moment the decision commits. It arises when the context a decision reads — balances, counters, session state, derived signals — lags behind or diverges from the events that have already happened. When the gap exceeds the decision's validity window, the system approves, blocks, prices, or routes based on a version of reality that no longer exists.
Updated
What is a context gap?
The term is part of the vocabulary Tacnode introduced to describe why automated decisions fail even when every component of the stack is working as designed. The requirement it names is that decision context be complete, consistent, and current — all relevant state in one place, one version of the present, reflecting reality right now; a context gap exists when the infrastructure cannot deliver all three simultaneously within the decision's window. Nothing is down, nothing is erroring — but the decision engine read state that was seconds behind the world, and in those seconds a concurrent withdrawal cleared, a fraud burst continued, or an agent took an action the policy check never saw.
Context gaps have two per-decision structural causes. The first is temporal — pipeline lag: derived state such as aggregates, velocity counts, and features takes pipeline time to compute, so it trails the events it summarizes. The second is spatial — inconsistent reads: context is split across specialized systems that cannot be read under one coherent snapshot, so a decision assembles inputs that reflect different moments. At organizational scale the failure takes a third form — the divergence problem, where many services' independently maintained views disagree with each other, not just with reality. And when the decision-maker is an AI agent, the same gap is harder to see: each step reads its own slightly different context, nothing errors, and the miss compounds across steps. Most production incidents blamed on "stale data" decompose into one or more of these.
The defining property is that a context gap is structural, not a tuning problem. Shorter cache TTLs, faster pipelines, and bigger clusters narrow individual lags without changing the architecture that produces the gap: state velocity plus concurrency means the world changes while context is being prepared and retrieved. The canonical treatment is the context gap page; the failure pattern under concurrent load is analyzed in context under concurrency.
How a context gap plays out
A card authorization illustrates the anatomy:
- 1.An account's spend velocity is computed by a stream job and cached; the pipeline lags events by about three seconds.
- 2.A fraudster fires twelve small transactions in two seconds across merchants.
- 3.Each concurrent authorization reads the cached velocity counter — which includes none of the burst, because the pipeline hasn't caught up.
- 4.Every individual check passes. The combined activity the full picture would have blocked has already committed.
No component failed. The gap between events and readable context — multiplied by concurrency — is the failure.
Why the context gap matters
The consequences scale with automation. When a human reviews a decision, staleness gets absorbed by judgment and delay; when the decision commits in fifty milliseconds with no human in the loop, the context gap converts directly into money: approvals that exceed limits, fraud that clears inside the pipeline lag, agents acting on state that changed mid-plan. The engineering question a context gap forces is not "how fresh is our data?" in the abstract, but "is the context this specific decision reads accurate within this decision's validity window?" — a per-decision budget, not a system-wide average.
FAQ
Related terms
The divergence problem: services each maintain their own pipeline over the same events, advancing at different paces, so their views of shared state disagree.
A validity window is the interval within which a decision's context remains an accurate basis for action — typically 10ms to 1s for automated decisions.
Decision coherence is the property that concurrent automated decisions act on the same version of shared state, instead of contradicting each other.
Derived context is state computed from events rather than stored directly — aggregates, velocity counts, features, embeddings — that decisions consume.
Cache coherence keeps multiple copies of shared data consistent across CPU caches or distributed services, so every reader sees the same value after a write.
State velocity is the rate at which the state a decision depends on changes. High velocity plus concurrency is what makes cached context structurally invalid.