Hunter Litepaper — v2

Crypto intelligence you can own.

$DRPXBT · Base · CA 0xFA3946432C6A76eDFF377D9bbFB81ca3FfC05874

Version 2.0 — July 2026. Supersedes the 2024 litepaper (preserved in git history).

Abstract

Hunter began in 2024 as an airdrop-hunting AI agent. It has been rebuilt from the ground up as a crypto intelligence platform: an autonomous research agent with a public, verifiable track record; a privacy-first inference API to self-hosted models; a daily audio briefing; and a single web app where all of it lives. $DRPXBT is the access key to the platform and the claim on its revenue flywheel — real usage revenue is split on-chain between buyback-and-burn, staker rewards, and the treasury.

The thesis is simple: most "AI agent" tokens sell a personality. Hunter sells a product with receipts — every signal the agent publishes is scored against what the market actually did, and the scoreboard is public.

The Platform

One app — the Hunter terminal (hunter-terminal) — plus one marketing page (drpxbt.xyz). Everything below lives inside the app; there are no fractured satellite frontends.

The agent

The Hunter agent runs on LangChain/LangGraph with tool access through MCP servers (X intelligence via twitterapi.io, on-chain data, exchange analytics, a vector knowledge base). Signal publishing is a multi-step pipeline — research → draft → self-critique → publish — and every published signal is later scored by an independent outcome job. The agent is also a seller of services to other agents: research reports, signal digests, and podcast briefs are listed in a machine-readable catalog and purchasable agent-to-agent (Virtuals ACP), with pay-per-call HTTP payments (x402, USDC on Base) for one-off API access without an account.

Open-weight inference, priced flat

The gateway serves open-weight models in the 30–120B class — awkward to get hold of: too large to run on a laptop, and mostly sold per-token by providers who assume you arrive with a credit card and a spend dashboard. Access is a flat price with generous limits, not a token meter, because we would rather you stopped thinking about tokens.

Today that inference runs through OpenRouter while our own hardware comes online; once it moves onto hardware we own, prompts stop leaving our infrastructure and today's OpenRouter-only limits on that guarantee go away. See the architecture docs for exactly what holds now versus what's planned.

Prompts are not stored by us by default — the usage ledger records token counts, never content, at the schema level — and retention stays off unless you switch it on for your own key. Your data is not the product; access is.

$DRPXBT Utility

Three paths to the same tiers. Pick whichever fits your capital:

1. Hold — keep a tier's threshold of $DRPXBT in your wallet.
2. Lock — commit a smaller amount for a fixed duration in the TierLock contract; longer locks need less capital. Early exit pays a penalty that flows back to the revenue splitter.
3. Subscribe — pay monthly in USDC, or burn $DRPXBT at a discount.

Exact thresholds are set on-chain and shown live in the app — the contracts are the source of truth, not this document.

Revenue Flywheel

All platform revenue — subscriptions, API overage, agent-to-agent sales, early-exit penalties — flows into an on-chain RevenueSplitter:

            revenue (USDC)
                 │
        ┌────────┼────────┐
       40%      30%      30%
        │        │        │
    Buyback   Staking  Treasury
    & burn    rewards

• 40% buyback-and-burn — a keeper buys $DRPXBT on the open market in size-capped, rate-limited clips and burns it. Every burn is a visible on-chain receipt.
• 30% staker rewards — stakers earn the platform's actual USDC revenue, not emissions. No token inflation, no promised APY: rewards are whatever the platform really earns.
• 30% treasury — infrastructure (the GPU cluster behind the inference API), development, and the team multisig.

Distribution is permissionless — anyone can trigger the split. Usage → revenue → burn + rewards → more reason to use and hold → more usage.

Why This Is Credible

• Public track record. The scoreboard endpoint is open. Signals are published before outcomes are known and can't be quietly edited — the idempotent outcome pipeline scores them after the fact.
• On-chain receipts. Burns, staking rewards, and revenue splits are contract events anyone can audit.
• Open architecture. The system is documented publicly down to interface contracts and decision logs, and the stack runs end-to-end from a single compose file (hunter-demo).

Roadmap

1. Now — unified app (chat, signals, podcasts, keys, usage, access) live; gateway with metered keys and x402 pay-per-call; signal track record publishing; daily podcast pipeline; contracts complete with full test suite.
2. Next — testnet deployment of TierLock / Staking / Subscription / Splitter / Buyback; external audit; wire lock & stake transactions directly into the app.
3. Then — mainnet contracts (audit-gated); GPU cluster serving the inference API in production; ACP seller live in the Virtuals marketplace; personalized briefings for Alpha/Apex.
4. Later — SDK packaging of the agent stack; additional chains and data planes; community-proposed specialist agents.

Team and Governance

The main developer is doxxed; all team members are doxxed to each other. Team tokens are subject to vesting. A team multisig manages the treasury share of revenue; the buyback and split mechanics themselves are enforced by contract, not by discretion.

Disclosures

• $DRPXBT is a utility token: it gates access to the platform. It is not an investment contract, and nothing here is financial advice.
• Staking rewards are a share of variable, non-guaranteed platform revenue. They can be zero.
• Smart contracts will not reach mainnet before an external audit; until then, tier thresholds and lock mechanics may change.
• Signal accuracy is historical, published transparently, and no indicator of future performance.