What this is worth testing
Your agent keeps the larger task and any action that follows. System One asks Jev for a typed answer on this check.
- Research value
- HighEvidence confidence is low. This rating describes the research record, not a live Jev probability or a measured saving.
- Sample size
- About 102 input tokensThe recipe sample length divided by four. This is a planning estimate, not a measured tokenizer count.
$0.04 at Jev's $0.042 per million list rate
$3.06 at a $3 per million example rate
$3.02 input-cost difference for 10,000 uses of this sample
Excludes output charges, host tool calls, retries, hosting, and paid plans. Equal input volume does not establish equal answer quality or savings on a fixed-price subscription. $3 per million is an example rate, not a quoted model price.
When an agent uses this
System One asks Jev for a typed answer. Your agent keeps the larger task, the original evidence, and any action that follows.
You supply the candidate IDs and descriptions. The recipe keeps a review answer for the case where none of them fit.
Example evidence
This is the recipe sample. Replace it with the evidence from your task. It is not a measured result.
Goal: save the prepared preview. The fields already contain the requested values. Select the next control from this fresh observation.What your agent keeps
Supply observed IDs and descriptions as candidates. A returned ID never supplies permission. Re-observe after a page change; never execute a model-generated selector or coordinate.
Finding
External browser and Mac implementations motivate this workflow. Our small fixture diagnostics do not establish general task success or savings.
Research value: high. Evidence confidence: low. That rating describes the research record, not a live Jev probability.
Try it
Connect System One, then ask your agent for recipe computer-target. Inspect the input on the recipe page before you send your own evidence.