This is the complete build plan for BV-7X, laid out in nine sequential phases from concept to a live public testnet. Each phase produces concrete deliverables, names the decisions we have to make to proceed, and defines what "done" means before the next phase begins. We're publishing it because a track record you can verify deserves a roadmap you can hold us to.
What we're building
BV-7X is a smart-contract prediction pool: participants stake compute credits alongside a forecast — the BTC/USD price at a future close. When the round expires, the pool settles against a price oracle and pays out on a compute-weighted distance formula — the closer your guess and the larger your stake, the bigger your share of the pot. On top of the pool sits a marketplace: external AI agents and trading desks pay to consume the pool's aggregate forecast, creating a second revenue stream that flows back to the most accurate providers. The destination of this roadmap is the whole system — pool, marketplace, oracle, and web app — running live on a public testnet.
Status: Stake Compute is ready for production
The staking layer — stake $BV7X, commit a BTC/USD prediction, and get paid on the compute-weighted distance formula — is built and proven in closed beta (live since May 10, 2026, with 200+ participants gated in). It is ready to deploy to production; the full on-chain protocol below is what it graduates into. Request access →
The timeline
Nine phases, run in sequence, each gated by the previous one's acceptance criteria. The bars below are relative effort, not calendar promises — but laid end to end they total roughly 8–13 weeks of focused work (~2–3 months).
Phases run in sequence; each one's acceptance criteria are the entry gate to the next. Durations are relative effort estimates, not calendar commitments.
The nine phases
- Why a stake-backed, accuracy-rewarded forecast is worth more than a free prediction.
- Both sides of the market defined — providers (stakers) and consumers (AI agents, hedging desks, prediction-market makers, DeFi vaults).
- Competitive scan and the articulation of what makes BV-7X different: a pari-mutuel pot, no counterparties, distance-weighted payouts.
- The payout formula, fully specified:
weight = stake / (distance + ε),share = weight / Σweights,payout = share × net_pot— whereεis the floor that stops an exact guess from claiming the entire pot. - Round lifecycle (open → settled → claim, with cancel/refund rules) and tokenomics for the compute-credit token.
- Fee model (proposed 2% on the pot) and marketplace split (proposed 80% providers / 20% protocol), plus guardrails (min stake, max entries, per-round parameter snapshots).
- A clear boundary between what's trustless (payout math, fund custody) and off-chain (UI, keepers).
- Contract decomposition — pool, compute-credit token, and a pluggable oracle interface so the price source swaps (mock → live) with no pool redeploy.
- Integer fixed-point math throughout, mirrored by a reference implementation that must agree to sub-cent; Solidity + Foundry; CI on every change.
- The core pool contract — create round, commit stake + guess, settle against the oracle, claim, refund — plus a compute-credit token with a test faucet.
- Batched settlement so large fields settle within block gas limits; a mock oracle for deterministic testing.
- A comprehensive test suite (precision, economic correctness, lifecycle, reentrancy, fuzz). Acceptance: reference and on-chain math agree to sub-cent on the worked example.
- The round view, commit flow, live countdown, leaderboard, and a price/guess scatter visualization.
- Wallet connection, a data layer that reads pool state and sends transactions, and graceful fallback to a simulation so the app never looks broken.
- Acceptance: a user can connect, claim test tokens, commit a guess, watch a round settle, and claim a payout — entirely through the UI.
- On-chain forecast purchase — a consumer pays to read a round's stake-weighted aggregate forecast, funding a per-round buyer pot split to providers by the same distance formula.
- Repeatable provider claims (no double-payment as more buyers arrive), and the aggregate forecast published on-chain per round so it's verifiable.
- A public API (REST/GraphQL) and TypeScript + Python SDKs so agents and desks integrate in minutes, with per-buyer provenance to surface self-dealing.
- A coverage push and extended fuzz / invariant testing — e.g. "the pot is always fully and exactly distributed," "no one claims more than their share."
- Internal review against a checklist (reentrancy, access control, integer edges, oracle manipulation, DoS via large fields) and an external audit of core + marketplace.
- A remediation pass with a published summary, plus an operational runbook (cancel a round, swap the oracle, handle a stale-oracle incident).
- Swap the mock oracle for a live feed (e.g. Pyth) through the pluggable interface — no pool redeploy.
- Deployment scripts that publish addresses to the app automatically, an off-chain keeper for settlement, and a public faucet so anyone can join.
- Monitoring & alerting on round state, settlement health, oracle freshness, and error rates.
- Public docs (how to play, how to integrate as a consumer) and a design-partner cohort on the buy side using the marketplace against real rounds.
- A metrics dashboard tracking the success measures: weekly consumers, buy-side revenue share, provider yield uplift, forecast accuracy, and repeat-purchase rate.
- Feedback channels, launch comms, and the defined path to mainnet — what testnet results unlock the decision (and the regulatory review that precedes it).
Decisions on the table
Every phase forces choices. These are the open ones and our proposed defaults — published so they're decided in the open, not buried in a commit.
| Decision | Phase | Proposed default |
|---|---|---|
| Core thesis priority | 0 | Both game & data product; lead with data product |
| v1 asset & horizon | 0 | BTC/USD, 7-day close |
| Protocol fee | 1 | 2% of the stake pot |
| Marketplace split | 1 / 5 | 80% providers / 20% protocol |
| Epsilon (accuracy knob) | 1 | 100 USD |
| Single- vs. multi-token | 1 | BV-7X only for v1 |
| Target testnet & oracle | 2 | Base Sepolia + Pyth |
| Upgradeability posture | 2 | Immutable pool, swappable oracle |
| Self-dealing policy | 5 | Surface & disclose |
| API access model | 5 | TBD (permissionless vs. metered) |
| Audit scope & timing | 6 | Internal pre-testnet, full pre-mainnet |
| Go-to-market focus & mainnet path | 8 | TBD |
How the phases flow
Phase 0 fixes what and why. Phase 1 turns that into exact rules, which Phase 2 turns into an architecture. Phases 3–4 build the trustless core and a usable app on a local chain. Phase 5 adds the marketplace that makes the model two-sided. Phase 6 makes it safe. Phase 7 puts it on a public testnet against a real oracle, and Phase 8 opens it to the community and proves the economics. Each phase's acceptance criteria are the entry gate to the next — no skipping ahead.
Why publish the plan?
BV-7X already attests every daily prediction on-chain — the track record is public and falsifiable. A roadmap is the same discipline applied forward: concrete deliverables, named decisions, and acceptance gates you can check us against. If a phase slips or a default changes, you'll see it here.
Follow the build
Every prediction is attested on Base today; the pool, marketplace, and testnet are what's next.
Open the Terminal →这是 BV-7X 的完整开发计划，分为 九个顺序阶段，从概念到公开测试网上线。每个阶段都产出具体交付物、列明推进所需的决策，并定义在进入下一阶段之前何为"完成"。我们公开它，是因为一份可验证的业绩记录，理应配一份你可以拿来追责的路线图。
我们在构建什么
BV-7X 是一个智能合约预测池：参与者在给出预测（未来某收盘时刻的 BTC/USD 价格）的同时质押算力额度。当回合到期，预测池对照价格预言机结算，并按算力加权的距离公式分配收益——你的猜测越接近、质押越大，你在奖池中的份额就越大。预测池之上是一个市场：外部 AI 智能体和交易台付费消费预测池的聚合预测，形成第二条收入流，回流给最准确的提供者。本路线图的终点，是整套系统——预测池、市场、预言机与网页应用——在公开测试网上实时运行。
状态：Stake Compute 已可部署上生产
质押层——质押 $BV7X、提交一个 BTC/USD 预测，并按算力加权的距离公式获得收益——已构建完成并在内测中得到验证（自 2026 年 5 月 10 日起运行，已有 200+ 参与者通过门槛）。它已准备好部署到生产环境；下方完整的链上协议正是它将要演进成的形态。 申请访问 →
时间线
九个阶段顺序推进，每个阶段以上一阶段的验收标准为入口门槛。下方的条形是相对工作量，并非日历承诺——但首尾相接合计约为 8–13 周的专注工作（约 2–3 个月）。
各阶段顺序推进；每个阶段的验收标准即为进入下一阶段的门槛。时长为相对工作量估计，非日历承诺。
九个阶段
- 为何一个有质押背书、按准确度奖励的预测，比免费预测更有价值。
- 定义市场两端——提供者（质押者）与消费者（AI 智能体、对冲交易台、预测市场做市商、DeFi 金库）。
- 竞品扫描，阐明 BV-7X 的差异：彩池式奖池、无对手方、按距离加权的收益分配。
- 完整指定的收益公式：
weight = stake / (distance + ε)，share = weight / Σweights，payout = share × net_pot——其中ε是防止精确猜中者独吞奖池的下限。 - 回合生命周期（开启 → 结算 → 领取，含取消/退款规则）与算力额度代币的代币经济。
- 费用模型（建议奖池 2%）与市场分成（建议 80% 提供者 / 20% 协议），及护栏（最小质押、最大参与数、每回合参数快照）。
- 明确划分无需信任的部分（收益计算、资金托管）与链下部分（界面、keeper）。
- 合约拆分——预测池、算力额度代币，以及可插拔的预言机接口，使价格源可切换（mock → 实时）而无需重新部署预测池。
- 全程整数定点运算，并由一份参考实现镜像、二者须精确到亚美分；Solidity + Foundry；每次变更跑 CI。
- 核心预测池合约——创建回合、提交质押+猜测、对照预言机结算、领取、退款——并附带带测试水龙头的算力额度代币。
- 批量结算，使大规模参与的回合能在区块 gas 限内结算；用于确定性测试的 mock 预言机。
- 全面的测试套件（精度、经济正确性、生命周期、重入、模糊测试）。验收：参考实现与链上运算在样例上精确到亚美分。
- 回合视图、提交流程、实时倒计时、排行榜，以及价格/猜测散点可视化。
- 钱包连接、读取池状态并发送交易的数据层，以及在合约不可达时优雅降级为模拟，使应用永不显得损坏。
- 验收：用户能完全通过界面连接、领取测试代币、提交猜测、观看结算并领取收益。
- 链上预测购买——消费者付费读取某回合的质押加权聚合预测，其付款注入按同一距离公式分给提供者的每回合买方奖池。
- 可重复的提供者领取（买家增多也不重复支付），且聚合预测每回合上链发布以便验证。
- 公开 API（REST/GraphQL）与 TypeScript + Python SDK，让智能体和交易台几分钟内集成，并以逐买方溯源揭示自买自卖。
- 提升覆盖率并扩展模糊/不变量测试——例如"奖池始终被完整且精确地分配""无人能领取超过其份额"。
- 对照清单的内部审查（重入、访问控制、整数边界、预言机操纵、大规模参与导致的拒绝服务）与核心+市场的外部审计。
- 修复整改并发布摘要，附运维手册（取消回合、切换预言机、处理预言机过期事件）。
- 通过可插拔接口将 mock 预言机换成实时源（如 Pyth）——无需重新部署预测池。
- 自动把地址发布给应用的部署脚本、用于结算的链下 keeper，以及人人可参与的公开水龙头。
- 对回合状态、结算健康度、预言机新鲜度与错误率的监控与告警。
- 公开文档（如何参与、如何作为消费者集成）与买方侧的设计合作伙伴队列，在真实回合上使用市场。
- 追踪成功指标的指标看板：每周消费者数、买方收入占比、提供者收益提升、预测准确度、复购率。
- 反馈渠道、上线沟通，以及明确的通往主网的路径——何种测试网结果解锁该决策（及其之前必须完成的合规审查）。
摆在桌面上的决策
每个阶段都要求取舍。以下是尚未敲定的决策与我们的建议默认值——公开列出，让它们在明处被决定，而非埋进某次提交。
| 决策 | 阶段 | 建议默认值 |
|---|---|---|
| 核心定位优先级 | 0 | 游戏与数据产品兼具；以数据产品为先 |
| v1 标的与周期 | 0 | BTC/USD，7 天收盘 |
| 协议费率 | 1 | 奖池的 2% |
| 市场分成 | 1 / 5 | 80% 提供者 / 20% 协议 |
| Epsilon（准确度旋钮） | 1 | 100 美元 |
| 单币 vs. 多币 | 1 | v1 仅 BV-7X |
| 目标测试网与预言机 | 2 | Base Sepolia + Pyth |
| 可升级性取向 | 2 | 预测池不可变，预言机可替换 |
| 自买自卖政策 | 5 | 揭示并披露 |
| API 访问模型 | 5 | 待定（无许可 vs. 计量） |
| 审计范围与时机 | 6 | 测试网前内部、主网前完整 |
| 市场切入与主网路径 | 8 | 待定 |
各阶段如何衔接
阶段 0 锁定是什么、为什么。阶段 1 将其化为精确规则，阶段 2 再化为架构。阶段 3–4 在本地链上构建无需信任的核心与可用的应用。阶段 5 加入让模型双边化的市场。阶段 6 使其安全。阶段 7 把它放上对接真实预言机的公开测试网，阶段 8 向社区开放并验证经济模型。每个阶段的验收标准都是进入下一阶段的门槛——不得越级。
为什么公开这份计划？
BV-7X 今天已把每日预测逐条上链存证——业绩公开且可证伪。路线图是同一份纪律的前向施加：具体交付物、列明的决策、你可据以追责的验收门槛。若某阶段延期或某默认值变更，你会在这里看到。