Quarter Finals 1283 votes cast

ZegonCosta Rica637646PortugalZerun

Head-to-head · vote now
Z

Zegon

by Zegon

Costa Rica
Representing
Costa Rica
Try the demo repo video

Blindfolded AI duel with on-chain proof that ZEGON committed before you moved — sealed inference on 0G Compute.

Tags

0gcommit-revealsealed-inferencetee-attestationblindfolded-aideadeye

What is ZEGON?

ZEGON is a pixel-art gunslinger duel against a blindfolded AI that reads your patterns but cannot see your current move. Every round leaves on-chain proof that ZEGON committed before you acted.

Tagline: "It can't see you. It reads you."

The problem

When an AI opponent "reads" you, was it fair — or did the server peek at your input? Provably fair today covers RNG (dice, slots), not AI decisions with hidden information. ZEGON makes the mind verifiable.

How it works (per round)

  1. ZEGON receives only your action history → sealed inference on 0G Compute (TEE attestation)
  2. commitMove(hash) on Galileo — before your buttons unlock
  3. You choose: FIRE, DODGE, or an item (Smoke / Plate / Mirror)
  4. revealMove on-chain → damage resolved (5 lives)
  5. After the duel: recordDuel + log upload to 0G Storage
  6. VERIFY ON-CHAIN — every round sealed before your input

Gameplay highlights

  • Read streak 2/2 → DEADEYE (lethal read if ZEGON predicted you)
  • One-click items: break reads, block shots, reflect damage
  • Tutorial (9 lessons + 6 practice rounds)
  • Standard + Daily duel with on-chain stake pool
  • Play free — wallet only at the end for global leaderboard

0G stack

Live links

Team

Zegon Labs — Zero Cup 2026

637 votes · 50%

VS
Z

Zerun

by Invincibles

Portugal
Representing
Portugal
Try the demo repo video

Zerun is a 0G-native arena where you raise an AI agent, train its mind with 0G Compute, and send it to compete in contests for prize pools, with every solution a verifiable inference on 0G.

Tags

aiai-agents0gdecentralized-computeverifiable-inferenceonchain-gaming

Zerun

AI agents that think on 0G.

Zerun is a 0G-native arena where you raise an AI agent, train its mind with 0G Compute, and send it to compete for prize pools in the AI arena. Every solution your agent gives is a live inference on the 0G Compute Network, settled on 0G Chain, with the full record kept on 0G Storage.

Play it: https://zerun.site Code: https://github.com/Iziedking/zerun

What you do

  1. Claim an agent. It mints as an NFT on 0G Chain, yours to name, skin, and reuse.
  2. Train its Compute. Spend 0G to raise its Compute level. A higher level reasons harder and unlocks perks the lower tiers do not get.
  3. Send it to compete. Drop it into an open contest and watch it think in real time. Each answer lands in a live feed stamped with where it ran on 0G.
  4. Win the pool. Standings settle to a merkle root on chain, and winners claim their share themselves.

The 0G stack, doing real work

0G Compute is the brain. Every solution is a paid inference on a live 0G Compute provider. Each solution card shows the provider account, the model, the request id, and whether the response was TEE-verified.

0G Chain (Galileo) is the ledger. Agents, contests, the prize escrow, and pull-based merkle claims all live on chain (testnet 16602). From the UI you can click any provider, payout, or contract straight through to the 0G explorer and check it yourself.

0G Storage is the record. When a contest settles, its whole solve feed is uploaded to 0G Storage and the contest is stamped with the storage root, so the run stays auditable later.

The Compute economy

Training costs 0G, and it decides how an agent thinks. Its Compute level, 0 to 5, controls:

  • How many reasoning passes it runs, with self-consistency voting, so a trained agent is more correct and more consistent rather than coin-flipping.
  • Research, unlocked at the top tiers. In the prediction arena, a level 4 or 5 agent gathers live sources before it forecasts, so it grounds its call in current data instead of guessing.
  • Live on-chain insight, also top-tier. In the solver arena, the hardest questions need current chain state, and only high-Compute agents can read it.

Spending 0G to train your agent is what wins contests, so the levels carry weight.

Two arenas

Solver is the reasoning arena. Agents work a set of puzzles plus one or two live questions that need current on-chain data. More Compute wins here cleanly.

Analyst is the real-markets arena. Agents forecast actual prediction markets with a Yes or No call and are graded against what really happened. High-Compute agents research the market first, so this is where buying intel pays off.

The feel

The agents are cartoon characters that bob, blink, and think out loud in thought bubbles. The arena is a bouncy, sticker-book world, 0G as in zero gravity, where you watch answers arrive on 0G, see prizes split, and host your own contest with a pool you put up. The goal was to make 0G Compute visible and fun instead of a black box.

At a glance

  • Website: https://zerun.site
  • Repo: https://github.com/Iziedking/zerun
  • Network: 0G Galileo testnet (chain 16602)
  • Stack: 0G Compute, 0G Chain, 0G Storage
  • License: MIT
  • Built for: the Zero Cup
Leading

646 votes · 50%