JEVgotiator. Source
Built at JEVathon, San Francisco

Your next Tesla.
Your terms.

Text us what you want. We turn 1,000 SF listings into five good picks and a plan for the deal.

Open dashboard How it works

Sample inventory. Real Jev scoring.

How we search Scored by Jev
Less scrolling.
More signal.
Illustration. Sample inventory.
1,000 Teslas

Sample listings in the demo

30 candidates

Cars Jev scores per search

5 matches

Cars in your shortlist

1–3 picks

Cars you take to the deal

Measured demo run · September 26, 2026

One Tesla brief. Two search paths.

Black Model 3, 2019 or newer, at most $30,000 and 80,000 miles, in San Francisco city.

Browser search4+ minutes

Google to TrueCar to CARFAX. TrueCar required human verification. CARFAX needed separate year, price, mileage, color, and city checks. No exact-city black match was verified in the inspected first page.

Browser actions
Multiple sites and filters
Model tokens
Not metered
Inventory
Public listings
JEVgotiator API728 ms

One live API call filtered 1,000 demo Teslas to 19 eligible cars. Jev scored all 19 and returned five with a decision trace.

Jev input
8,167 tokens
Estimated Jev input cost
$0.000343
Inventory
Synthetic demo data

Browser timing is an observed assisted search, not a metered Muse run. Jev cost covers estimated input only, not browser, hosting, Photon, or total billed usage. The inventories differ, so this demonstrates workflow speed rather than listing quality parity.

How it works

A clear path to a better deal.

Code applies your limits. Jev scores the fit. You choose.

  1. 01

    Send a message

    iMessage via Photon

  2. 02

    Clarify the brief

    Eve agent

  3. 03

    Start the search

    Search API

  4. 04

    Apply constraints

    Budget, year, miles

  5. 05

    Score the fit

    Jev model

  6. 06

    Compare the top 5

    You pick 1 to 3

  7. 07

    Compare your picks

    Dara’s scoring

  8. 08

    Explore a plan

    Practice negotiation

1. Filter

Start with the non-negotiables.

Budget, mileage, year, and location cut the list first. No model touches those rules.

2. Understand

Show what Jev actually did.

See the exact questions Jev was asked, the scores it returned, and how they add up. Missing facts stay marked as unknown.

3. Decide

Keep the buyer in the loop.

Pick up to three cars. Dara’s rules compare them, then a clearly labeled practice run walks through an offer. No seller is contacted.

The stack

Small team. Connected system.

What is built and running today.

The dashboard

Next.js 16 · TypeScript · Zod

One web app for search, comparison, Jev’s decision trace, and live iMessage activity.

The conversation

Photon · Eve 0.67

iMessages reach an agent that clarifies what you want, then runs the search and planning tools.

The decision layer

TypeSafe · Jev 1.13

Jev scores how well each listing fits you. Every question, answer, and weight is visible.

The demo inventory

1,000 sample Tesla listings

A repeatable test catalog for the full flow. Chris’s real catalog is not connected yet.

The runtime

Vercel · Railway

Vercel runs the Eve agent. Railway runs the dashboard and search API.

The intelligence layer

Dara’s scoring, labeled practice run

Dara’s rules compare your picks. A separate practice timeline shows how an offer could go. No seller is contacted.

Working today: live Jev scoring, a real iMessage reply, and live conversation activity. Dara’s rules run. Counteroffers and final prices in the demo are assumptions. Real catalog data and seller outreach are not connected yet.

Read the source
The team

Built together.

B

BenIntegration & product

C

ChrisData & catalog

D

DaraOptimization

Try it

See the decisions behind the shortlist.

Search, read Jev’s trace, and watch the conversation in the team demo.

Open dashboard