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 moderate. This rating describes the research record, not a live Jev probability or a measured saving.
- Sample size
- About 115 input tokensThe recipe sample length divided by four. This is a planning estimate, not a measured tokenizer count.
$0.05 at Jev's $0.042 per million list rate
$3.45 at a $3 per million example rate
$3.40 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.
The service is unreachable, but its processes are healthy. Tests must be read-only.What your agent keeps
The agent proposes hypotheses and executes read-only checks. A missing correct hypothesis cannot be recovered by ranking.
Finding
The SRE study measured more passing attempts overall, with regressions. Trace attribution also showed low joint diagnosis accuracy.
Research value: high. Evidence confidence: moderate. That rating describes the research record, not a live Jev probability.
Try it
Connect System One, then ask your agent for recipe diagnostic-test. Inspect the input on the recipe page before you send your own evidence.