Anchor launches Action Index to replay proven web workflows
The AMW Read
The reported workflow-reuse results meaningfully update agent cost-performance expectations, but undisclosed evaluation details limit conclusions about broader market impact.
Anchor launches Action Index to replay proven web workflows
Anchor launched Action Index, a system that stores shared execution memory and replays proven web workflows as deterministic code. The reported approach lets computer-use agents reuse successful execution paths. According to the supplied report, a small model using Action Index matched a frontier model's success rate across 300 tests while reducing token costs by 100 times. These are reported test results; the supplied text does not identify the models, task mix, baseline configuration, or evaluation methodology.
The market significance lies in where agent performance comes from: the model itself or the execution system around it. Anchor's approach places reusable workflow knowledge in that surrounding system, making orchestration and execution memory potential sources of differentiation. If the reported results generalize, repeated web tasks could require less model reasoning while retaining comparable completion rates. That would strengthen the case for agent products whose value rests on reliable workflow execution. Token-cost savings alone, however, do not establish equivalent reductions in total operating costs.
For builders, the concrete implication is to evaluate workflow reuse alongside model selection. A useful deployment test would measure completion rates, token spending, and maintenance effort when web interfaces change or tasks depart from stored paths. For investors, the unanswered question is how broadly the reported advantage extends beyond the 300 tests: the commercial value depends on repeatable savings under customer conditions, with reliability preserved as workflows evolve.