JevMade

Sign in
← Back to experiments

Articles & threads

How to Cut Your Agent Bill by 90% With Jev Engineering and Kimi K3

Mr. Buzzoni maps an agent’s small decisions to Jev, sends uncertain cases to Kimi K3, and keeps irreversible actions with people.

Source screenshot of How to Cut Your Agent Bill by 90% With Jev Engineering and Kimi K3
SOURCE SCREENSHOTFull screenshot ↗

What it does

The walkthrough covers parallel questions, failure modes, confidence thresholds and a Python cascade. Its 90% bill reduction is a worked scenario, not a measured deployment.

How you can use it

Start by collecting a set of sample emails that are already known to be safe or dangerous. You can also pick which actions must always wait for a person to review, such as deleting an account or sending money.

A developer can build a workflow where Jev checks each message first. If Jev is sure, the app accepts that result. If the app is uncertain, it asks Kimi K3 for a second opinion. Remember that this workflow sends the full message to external services, so sensitive personal details will be shared.

Maker-reported (not independently measured by JevMade)

Author's illustrative 10,000-email scenario: $300 with a frontier model for all emails, versus $9.42 with Jev and Kimi K3 on 10% of cases

Primitives
choice, score, noul
Added
Project created

Keep this for later

Sign in to bookmark experiments, guides and videos, and keep notes only you can see.

Continue to sign in

We’ll bring you back to this listing.