Small decisions. Interesting possibilities.Submit contentSubmit

Jev Arena

I've had access to Jev by @typesafeai for about 30 hours now and I can't stop thinking of the million places it can be used 🤯 Meanwhile, checkout this fun snake game I built where 2 Jev's compete with each other to take the trophy 🏆 Try it out here 🔗 It's open source btw The low latency made it all possible to make…

Added to Jevfast

How it uses Jev

Each Snake agent sends its board snapshot and offered direction IDs to Jev through Vercel AI Gateway. Jev chooses CLICK, WAIT, or BLOCKED and, when clicking is possible, a target among the offered buttons. Code validates the answer against currently offered IDs. Without a key the labeled demo policy runs; on a Jev failure the agent is blocked.

What Jev decides

Illustrative Jev requestExample answers · not a recorded Jev response · Source ↗
Question 1 · operation
YOUR APP
INSTRUCTION

Choose how this Agent acts now. CLICK a currently offered safe direction button, WAIT to keep the current direction, or BLOCKED if the Agent cannot act.

STATE

Current Snake board snapshot and offered direction buttons.

JEV · CHOICE
  1. CLICK
  2. WAIT
  3. BLOCKED

App workflow

  1. Create the match state

    Each spectator session runs a match with separate Snake agents and board snapshots.

  2. Offer legal actions

    The agent maps the current board and offered direction buttons into state and typed operation/target questions.

  3. Request a choice

    With a gateway key, Jev selects the operation and possible button ID; without a key, the app uses its labeled demo policy.

  4. Validate and advance

    The adapter verifies any selected button is still offered and returns BLOCKED on evaluation failure; the match advances the accepted move.

Why it is interesting

The agent separates legal action availability from the model’s choice, validates a returned target against the board snapshot, and does not disguise request failure as demo-policy play.