Glossary · Cost engine

Deterministic replay

Re-running a compiled skill as a fixed sequence of indexed actions, with zero LLM calls, so the same goal produces the same path every time.

Start free All terms

What is deterministic replay?

Deterministic replay executes a previously compiled skill as an explicit, ordered program of browser actions keyed to indexed DOM elements, without invoking a language model. Because the plan is fixed and re-resolved against the live page rather than regenerated, the same goal yields the same action path — fast, cheap, and auditable.

Why it matters

Replay is what makes a cache hit nearly free: once a skill is compiled, re-running it costs browser time, not model tokens. It also makes runs reproducible and reviewable, which matters for RPA and authorized testing where you need the same flow to behave identically across thousands of executions.

See it in context: read how Twin compiles and replays a run, follow the cost-cutting guide, or browse use cases and comparisons.

Related terms

Skill compilation

Turning a successful agent run into a reusable, replayable skill — a compact program of indexed actions you can call again without re-planning.

Semantic dispatch cache

A vector cache that matches a new, re-phrased request to a skill you already compiled — so a repeated task skips the LLM.

DOM indexed state

A compact, numerically-indexed map of a page’s interactive elements that an agent acts on by index instead of raw HTML.

Marginal cost per run

The cost of one more execution of a workflow. On most browser infra it’s flat; on Twin it falls as cache hits replace LLM calls.

Back to the full glossary

Run your first skill

Give an LLM agent a real browser, compile the workflow once, and watch the marginal cost fall as the cache takes over.

Start free See pricing