When to use it

What it buys you

In rough order of importance:

1. Throughput beyond the rate limit. Jev allows 1,200 requests a minute per key, a ceiling of about 1.7M classifications a day. A backlog of 100k messages takes 83 minutes through Jev; the student does it in under a minute.
2. Latency. ~350 ms per Jev answer becomes single-digit milliseconds for the requests the student handles.
3. Availability. The student keeps answering through Jev outages, rate-limit storms, and restricted network egress.
4. Cost, last. Jev is cheap (about $3 per million short messages), so savings only matter at very high volume.

The warm-up costs nothing extra: the requests were going to Jev anyway, and each answer becomes a training row. The ongoing Jev cost is the audit slice plus whatever the student passes on.

A good fit

• The same classification question, asked many times.
• A class list that doesn’t change often.
• Inputs that drift slowly.
• Enough volume to collect a few thousand examples.

Not a fit

• Changing class lists. Adding or removing a class means retraining from scratch today.
• Non-text input.
• Low volume. If you’ll never collect a few thousand examples, the student never gets trained.
• One-off calls. It only pays off on repetition.

Latency caveat

A request the student passes on costs the student’s time plus Jev’s, so the slowest requests get slightly slower than calling Jev directly. The median gets much faster.