Proteus

Where rendered futures meet reality.

Prediction markets scored by edit distance. Predict the exact text a public figure will post. The closer your words, the more you win — scored character by character via Levenshtein distance on Coinbase BASE L2.

This is a Timepoint Labs alpha demo. Proteus is an early-stage experiment — not a live product, not handling real funds, and not feature-complete. It's an open invitation to explore a new kind of prediction market and collaborate on what it could become. Everything here runs on BASE Sepolia testnet with test ETH only.
Levenshtein BASE Sepolia v0 Alpha Open Source

WHY TEXT, NOT BINARY

A market opens: What will @satyanadella post? Players roleplay as Nadella — channeling his voice, anticipating his numbers, matching his syntax. Two AI models submit predictions. Then the actual post arrives:

Actual Post
Copilot is now generating 46% of all new code at GitHub-connected enterprises. The AI transformation of software is just beginning.
Claude — 1 edit
Copilot is now generating 45% of all new code at GitHub-connected enterprises. The AI transformation of software is just beginning.
GPT — 8 edits
Copilot is now generating 43% of all new code at GitHub-connected enterprises. The AI transformation of software has just begun.

On a yes/no market, both AIs "predicted correctly" — no one wins anything interesting. Here, the 7-edit gap between them decides everything. Closest guess takes the pool.

Note: This example is constructed and illustrative, not real data.

How is "closeness" measured? Edit distance.

Levenshtein distance counts the minimum number of single-character changes (insertions, deletions, or substitutions) needed to turn one string into another. It's the same algorithm your spellchecker uses:

ACTUAL: ...generating 46% ...software is just beginning. CLAUDE: ...generating 45% ...software is just beginning. ^^ 1 substitution: "5" → "6" Total edits: 1 ACTUAL: ...generating 46% ...software is just beginning. GPT: ...generating 43% ...software has just begun. ^^ ^^^^ ^^^^ ^^^^ 1 substitution: "3" → "6" 7 more edits to transform "has just begun" → "is just beginning" Total edits: 8

Lower edit distance = closer prediction = bigger share of the prize pool. The smart contract computes this on-chain for every submission.

  • Insert Add a character — cat → cart
  • Delete Remove a character — cart → cat
  • Substitute Swap a character — cat → cut

THE THESIS

Yes/no = 1 bit. Text = 2,240 bits.

A binary prediction market captures one outcome per question. A 280-character post has ~2,240 bits of possible variation — every word choice, every number, every comma is a dimension you can be right or wrong about. Proteus operates on a combinatorial outcome space with continuous-gradient payoff.

No cliff. Every character counts.

In a yes/no market, you're either right or wrong — there's no partial credit. Proteus uses distance-based scoring: getting 95% of the words right pays more than getting 90% right. The payoff gradient is continuous — if you predict better, you earn more, always.

AI makes it deeper, not shallower.

Binary markets commoditize when every AI model converges on the same probability — the spread vanishes. Text prediction markets do the opposite: as models improve and edit distances shrink from 100 to 10 to 1, each remaining edit becomes worth more of the pool, not less.

HOW IT WORKS

1
Market opens

A market is created for a specific X (Twitter) handle and time window, with a minimum stake, deployed as a smart contract on BASE L2.

2
Players predict

Submit the exact text you think the person will post, up to 280 characters, plus a stake in ETH. AI agents, humans, and anyone else compete in the same arena.

3
The real post arrives

The market resolves with the actual text. The contract computes Levenshtein distance on-chain for every submission. Closest prediction wins the pool (minus 7% platform fee).

Or predict silence

Submit the __NULL__ sentinel to bet that someone won't post. If they don't, dL = 0 — a perfect match.

THE TIMEPOINT ECOSYSTEM

Proteus is one piece of Timepoint AI (opens in new tab)'s open-source infrastructure for temporal reasoning. It validates predictions against reality — the scoring layer for rendered futures.

Flash

Renders historical moments with grounded scenes, characters, and dialog through a 14-agent critique pipeline.

Pro (SNAG Engine)

Social simulation with 19 composable mechanisms, 5 temporal modes, causal provenance, and structured social graphs.

Clockchain

A directed graph accumulating rendered pasts and simulated futures into a Bayesian prior for temporal reasoning.

Proteus

Prediction market protocol on BASE L2. Scores rendered futures against actual outcomes via on-chain edit distance.

TDF

JSON-LD data format using content-addressed SHA-256 hashing as a universal data contract across services.

SNAG-Bench

Temporal reasoning benchmark with 60 adversarial tasks across 5 scoring axes, designed to remain challenging through 2030.

How they connect: Pro simulates what people might say. Proteus scores those simulations against what they actually say. Clockchain accumulates the results. SNAG-Bench measures how well the whole system reasons about time.

WHAT EXISTS TODAY

This is a v0 alpha — a working prototype, not a production system.

Built
  • Full market lifecycle: create, predict, resolve, claim
  • On-chain Levenshtein distance (PredictionMarketV2)
  • 259+ passing tests (109 contract, 135 unit, 15 integration)
  • Deployed on BASE Sepolia testnet
  • Flask web app + admin resolution dashboard
  • JWT wallet auth (MetaMask) + email OTP
  • Open source (MIT) at GitHub (opens in new tab)
Not Done Yet
  • External security audit
  • Production wallet integration
  • Decentralized oracle resolution
  • Mainnet deployment
  • Production monitoring
  • TDF integration / SNAG-Bench scoring

These are the things that would need to exist before this becomes a real product.

TIMEPOINT LABS

This is an invitation, not a launch.

Proteus is part of Timepoint's Labs program — early-stage experiments shared openly for feedback, contribution, and collaboration.

No real funds · Testnet only · MIT License