jevtrafficsim
Chicago traffic simulator comparing fixed, adaptive and Jev-guided citywide signal policy on identical scenarios.
Overview
Jev Traffic Sim is an interactive Chicago traffic simulator built around one experiment. You watch a single car cross real Chicago geography while the whole city simulates around it, and the same scenario is played headlessly by two deterministic baselines: fixed-rotation signals and adaptive signals reacting to local queues. In the third condition, Jev, reached through the Vercel AI Gateway, sees only citywide traffic state and answers bounded typed questions whose answers become corridor and region weights (0.5 to 2), a pressure scale (0.5 to 1.5) and a coarse hint.
Jev never sets a signal state. The existing controller translates its policy into legal local decisions, and phase order, minimum and maximum green, yellow, all-red and starvation protection are enforced in code regardless of the model. Fairness is enforced by hashing each scenario into a fingerprint, sharing demand and incident scripts across runs, marking runs a human touched as non-comparable, and never including the watched car's identity or route in requests, which three tests check. No winner score is computed.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Score | How much weight should each corridor or region get in signal policy?(inferred) | — |
| Score | How strongly should controllers respond to queue pressure?(inferred) | — |
| Choice | Which coarse policy hint applies to current citywide traffic?(inferred) | — |
How it works
- Scenario (trip, traffic, driver, seed) is fingerprinted
- Fixed and adaptive baselines run the same scenario headlessly
- Jev receives citywide state and returns bounded policy values
- Values are validated against hard limits and turned into local signal decisions
- Results are shown side by side only for comparable runs
Strengths
- Strict same-scenario fairness rules
- Safety constraints enforced in code, not by the model
- No privileged information about the watched vehicle
Caveats
- Simulation only
- Jev only adjusts weights, not signals directly
- No aggregate result claims in the README
At a glance
- Type
- Demo
- Status
- prototype
- Language
- TypeScript
- Runtime
- browser simulation with worker
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
TypeSafe AI's first model Jev takes on an entire city's traffic.
hellogumbo/awesome-jev
Frequently asked questions
Does Jev control traffic lights?
No. It returns bounded corridor weights and a pressure scale that local controllers turn into legal signal changes.
How is fairness ensured?
All three conditions share one scenario fingerprint, demand list and incident script.
Who wins?
No winner score is computed; the panel shows real run fields only.
Related projects
Real-time bot battles where you write a fighter's strategy in plain English and a System One model pilots it several times a second.
ProjectbetaTypeScriptBot that plays Red Alert 2 web skirmishes, with code handling mechanics and a decision model making stance, threat and timing calls.
ServicebetaPythonMultiplayer 3D science-fantasy version of Houston's River Oaks District with Jev-powered NPC companion, chauffeur and smart drive.
ProjectbetaTypeScriptBrowser 3D planetary rover sandbox where a TypeSafe controller chooses the next action from complete candidates, versus a baseline.
ProjectbetaTypeScriptKorean RL tycoon game training a company decision model that answers situations with choice, probabilities and confidence.
ProjectprototypePythonDoom runs as a spacecraft payload over real flight software; Jev pilots from the ground every half second while Claude reviews.
ProjectprototypePython