20:"$Sreact.view_transition" 21:I[9754,["/_next/static/chunks/2t36f5p4pjzlb.js","/_next/static/chunks/1gyqeidjm5eoa.js","/_next/static/chunks/0_fjnxc4tt_l_.js","/_next/static/chunks/1ymt1shyhmu-4.js","/_next/static/chunks/3kv2zf0i7s2wc.js","/_next/static/chunks/1vs_3g_g75zmk.js","/_next/static/chunks/2g4dh802s8oz3.js","/_next/static/chunks/1bc3f0shukf1j.js","/_next/static/chunks/192kgs9v_t14j.js","/_next/static/chunks/0z6pw0v5c_lx5.js","/_next/static/chunks/08r60e6icmtkc.js","/_next/static/chunks/3trx6c0i7_x-3.js","/_next/static/chunks/40sro5bzp9bi_.js","/_next/static/chunks/1wgf3tk38y6cf.js"],"ProjectDetail"] 22:T6682, vibe-investing · AI 投资委员会

AI 驱动的投资研究与量化工具集

让每个投资判断，
都要经过一轮委员会表决。

vibe-investing 把 NASDAQ、S&P500 与加密货币放进同一张会议桌——用多个 LLM 智能体交叉回测、核验论据，再收敛成一个带置信度的判断。

进入研究库 GitHub · gameworkerkim/vibe-investing

启动方 gameworkerkim · 语言 HTML · 议题 NASDAQ · S&P500 · 加密货币

RAG·检索

E-01

回测·校验

E-04

情感·研读

E-07

策略·走查

E-02

对冲·否决

E-05

指南·审计

E-09

76%

通过

本轮议题 · 收敛置信度

附 1 条对冲否决注记

▲ mNAV 套利入线
▲ BTC–Nasdaq 耦合
▼ 情绪信号被降权

议题 · LLM 量化工具收敛旁听席位 5 证据入线 9

表决现场

不是一个人的盘感，而是一场多智能体的表决。

美股、S&P500 与加密货币被放进同一条证据管线：每个 LLM 席位先核验自己那一份证据，再把结论汇到一个菱形令牌上，显示置信度与「通过 / 否决」。

证据入线

E-01 · RAG 检索 / E-07 · 情感研读 / E-09 · 指南审计

各席位带着自己的证据链进场，指向同一议题。

交叉核验

席位之间互相校核对方的论据，同一份数据由不同智能体反复读出。

结算置信度

收敛后的判断带一个置信度数字，落在中央令牌上。数值来自证据一致度，不是营销口号。

—— 议题清单 ——
● mNAV 套利价差
● BTC–Nasdaq 耦合强度
● Alpha Arena 多智能体竞技
● 自适应动量策略
○ 待定 · 单一来源信号

委员会席位

每一类内容，都是会议桌上的一席。

仓库里那些真实存在的量化条目，按角色各占一席：有的负责策略，有的负责核验工具，有的负责把安全与纪律放进表决。

席 01

自适应动量 · 策略席

面向韩国股票与 ETF 的自适应动量量化仪表盘，提供无需 API 密钥即可运行的模拟模式。

表决 · 通过（策略入线）

席 02

ARDS-X · 衰退判定席

用 FRED 与 yfinance 的真实数据，把市场下跌自动归入不同衰退类型并给出可视化仪表盘。

表决 · 附核验说明

席 03

LAON VaultGuard · 纪律席

跨平台多 LLM 安全审计工具，检测 Git 仓库里泄露的云私钥，支持离线运行。

表决 · 通过（纪律入线）

席 04

CASSANDRA AI · 警戒席

基于 LLM 的分析系统，针对 KOSDAQ 可疑公司披露进行判断，使用 DART 数据并做多 LLM 交叉验证。

表决 · 通过（交叉验证）

核验方法

判断之前，先核验证据来源。

这套工具集的价值不只是给出结论，更在于把结论的来路摆上台面——数据来自哪里、是否可复现、是否需要额外密钥。

来源核验

真实数据入线

衰退判定与回测以 FRED、yfinance、DART 等外部数据为输入，而非拍脑袋的假设。

证据 · 外部数据源

可复现性

模拟模式可跑

自适应动量仪表盘提供 mock 模式，不配密钥也能先跑通流程、验证逻辑。

约束 · 部分工具仍需 API

多智能体交叉

同一结论多重读出

多智能体回测与多 LLM 交叉验证，让单一智能体的偏误被对方席位检出。

机制 · multi-agent 交叉

反对票与对冲

容许反对票，才是成熟的投资委员会。

一件量化工具的价值，也在于它愿意把不能答的部分标注出来——而不是只发射高置信度的绿灯。

✕
研究性质，非投资建议

仓库里多为研究型项目，代码不应被当作稳赚的买卖信号。这是纪律席反复强调的边界。
◌
内容以韩文为主

大量指南与说明以韩文撰写；中文与英文使用者阅读时需留意语言门槛。
◌
部分工具需要密钥

真实数据场景可能需要你自己的 API 密钥或外部数据；模拟模式可作进入门槛。

委员会在每一条「通过」旁附上风险注记，而不是把反对票藏在统计里。 对冲状态：持有反对票以约束过度自信。

研究库入口

走进席位，开始你的核验。

从研究库首页出发，把 NASDAQ、S&P500 与加密货币的议题放进你自己的回测与判断管线。

vibequant.cc · 研究库

AI 投资研究与量化工具集的项目入口

→

gameworkerkim/vibe-investing

策略、指南、审计与分类器的源仓库

→

入会须知 · 三条边界

研究性质，非投资建议内容以韩文为主部分工具需 API 密钥

23:T65b9, vibe-investing · The AI Investment Committee

AI-powered investment research & quant tools

Every judgment
crosses a committee vote.

vibe-investing pulls NASDAQ, S&P500, and crypto onto one table — multi-agent backtesting, cross-checked evidence, and a decision that lands with a confidence number attached.

Enter the research library GitHub · gameworkerkim/vibe-investing

Maintained by gameworkerkim · Language HTML · In session NASDAQ · S&P500 · Crypto

RAG · fetch

E-01

Backtest · check

E-04

Sentiment · read

E-07

Strategy · walk

E-02

Hedge · veto

E-05

Guides · audit

E-09

76%

PASS

this session · converging confidence

1 dissenting hedge noted

▲ mNAV arbitrage in line
▲ BTC–Nasdaq coupling
▼ sentiment signal down-weighted

Issue · LLM quant tool convergence Observing seats 5 Evidence lines 9

In session

Not one trader's hunch — a multi-agent vote.

US equities, S&P500, and crypto share the same evidence pipeline: every LLM seat verifies its slice first, then the results land on a diamond token stamped with a confidence number and a pass / reject verdict.

Evidence enters the line

E-01 · RAG fetch / E-07 · sentiment read / E-09 · guides audit

Each seat walks in carrying its own chain of evidence, aimed at the same issue.

Cross-check

Seats challenge each other's premises; the same data is read more than once, by different agents.

Settle on confidence

The converging judgment carries a numeric confidence on the central token — driven by evidence agreement, not a slogan.

— on the docket —
● mNAV arbitrage spread
● BTC–Nasdaq coupling strength
● Alpha Arena multi-agent arena
● adaptive momentum strategy
○ tabled · single-source signal

The committee seats

Each body of work fills a chair.

The real quant entries in the repository map onto roles around the table — some run strategy, some verify, and some keep discipline inside every ballot.

Seat 01

Adaptive Momentum · strategy

An adaptive-momentum quant dashboard for Korean stocks and ETFs, with a mock mode that runs without an API key.

Vote · pass (strategy in line)

Seat 02

ARDS-X · downturn classification

Uses FRED and yfinance data to sort market drawdowns into decline types, surfaced in a visualization dashboard.

Vote · verified with notes

Seat 03

LAON VaultGuard · discipline

A cross-platform multi-LLM security audit tool that detects leaked cloud keys in Git repositories, with offline support.

Vote · pass (discipline in line)

Seat 04

CASSANDRA AI · caution

An LLM-based analyst for suspicious disclosures in the Korean KOSDAQ market, using DART data with multi-LLM cross-validation.

Vote · pass (cross-validated)

How it verifies

Before deciding, audit the source.

The value here isn't just conclusions — it's that reasoning is put on the table: where data comes from, whether it reproduces, and which calls stay unresolved.

Source audit

Real data in the line

Downturn classification and backtests draw on outside sources like FRED, yfinance, and DART — not off-the-cuff assumptions.

Evidence · external data feeds

Reproducible

Mock mode that runs

The adaptive-momentum dashboard ships a mock mode, so the pipeline runs and the logic checks out before any key is configured.

Constraint · some tools still need keys

Multi-agent cross-check

Same call, several reads

Multi-agent backtesting and multi-LLM cross-validation let one agent's bias get caught by the chairs across from it.

Mechanism · multi-agent cross-check

Dissent & hedge

A grown-up committee welcomes the no vote.

A quant tool earns trust when it also flags what it cannot answer — not only when it keeps turning on a confident green.

✕
Research, not advice

Most entries are research projects, not signals to trade on. The discipline seat keeps re-stating that boundary.
◌
Mostly Korean content

A large share of the guides and notes is written in Korean — a real reading cost for other-language visitors.
◌
Some tools need keys

Live-data scenarios may want your own API keys or external feeds; mock mode is the faster way in.

Every "pass" is filed beside a risk note, not hidden in the tally. Hedge status: holding dissent to blunt overconfidence.

Research library

Take your seat, start verifying.

Leave the page and run NASDAQ, S&P500, and crypto issues through your own backtest and judgment pipeline.

vibequant.cc · research library

The project hub for AI investment research and quant tools

→

gameworkerkim/vibe-investing

Source repository of strategies, guides, audits, and classifiers

→

Before you sit in · three boundaries

Research, not investment advice Mostly Korean content Some tools need API keys

a:[["$","$20",null,{"enter":{"secondary-forward":"secondary-forward","secondary-back":"secondary-back-enter","default":"none"},"exit":{"secondary-forward":"secondary-forward","secondary-back":"secondary-back-exit","default":"none"},"update":{"secondary-forward":"secondary-forward","secondary-back":"secondary-back-update","default":"none"},"default":"none","children":["$","$L21","cmt2zh09s00no62714hgl9u44:zh-CN:beautified",{"project":{"id":"cmt2zh09s00no62714hgl9u44","slug":"gameworkerkim-vibe-investing","name":"vibe-investing","description":"AI投资研究、量化策略与工具集","authorName":"gameworkerkim","authorVerification":{"verified":false,"verifiedAt":null},"category":"development_tools","subcategory":"data_ai_tools","experienceUrl":"https://vibequant.cc/","qualityScore":83,"featured":false,"officialSite":null,"views":{"original":{"available":true},"official":{"available":false,"reason":"disabled_by_policy"},"beautified":{"available":true,"reason":null},"defaultView":"beautified"},"beautifiedView":{"revisionId":"cmt2zh0be00o062712yoajr6p","kind":"artifact","artifact":{"schemaVersion":"localized-static-html-v4","kind":"static_html","locales":["zh-CN","en"],"defaultLocale":"zh-CN","documents":{"zh-CN":"$22","en":"$23"},"designDirections":{"schemaVersion":"dodot-design-directions-v1","directions":[{"id":"dir-alpha-committee","name":"LLM 投资委员会表决","projectMetaphor":"多个 LLM 智能体组成一个投资委员会，分别核验论据并表决，最终收敛成一个带置信度的投资判断。","signatureVisual":"横向一排异形委员会席位（不同形状的节点代表不同 LLM），每个席位带一条证据链入线，汇聚到中央一枚表决令牌；令牌显示置信度百分比与\"通过/否决\"。","heroGeometry":"一组发散的证据链以不同角度汇入画面中心的一个菱形表决节点，节点四周标注当前议题、置信度与对冲符号。","sectionSequence":["表决现场","委员会席位","核验方法","反对票与对冲","研究库入口"],"surfaceModel":"深色会议厅舞台：每张席位是一块不规则的\"芯片茶几\"面板，表面带细密的证据纹线；交汇处令牌用高对比金属高光。","palette":"暗墨青底 + 琥珀表决色 + 冷白证据纹 + 单一玫红作为否决警示。","typography":"等宽衬线用于核对文字（象索引证编号），粗体无衬线用于议题标题，斜体小字用于唱票过程。","desktopComposition":"左右纵深式舞台：左列是席位与证据进线，右侧悬浮表决令牌与理由清单构成主视觉。","mobileTransformation":"席位列折叠为可顺序滑动的自定义 panel 组，令牌转为顶部全宽决策条。","mediaStrategy":"无本地截图素材，删除媒体区；用委员会席位+证据链线+令牌排版承担全部视觉。","forbiddenPatterns":["营销 Hero","三卡指标墙","浏览器/终端/仪表盘外壳","发光渐变粒子场"],"compositionFingerprint":"多源证据线汇入中央表决令牌","evidenceBasis":"仓库事实：多智能体回测、multi-agent、Alpha Arena、mNAV 套利、多 LLM 交叉验证"},{"id":"dir-beta-arb-edge","name":"套利价差波形","projectMetaphor":"把资产（mNAV、BTC-Nasdaq 耦合）的实时价差当作一条波形线，量化工具的价值就是抓住波边的信号而不是静态截图。","signatureVisual":"一条贯穿页面的连续价差波形线，线上标记出历史上被工具捕捉的高光点，波形下以半透明面积表达被识别的时间窗。","heroGeometry":"大跨度曲线从视口左缘切入，在中段形成可读的波峰波谷，线尾并入 CTA 的方向箭头。","sectionSequence":["波动中的机会","耦合信号","套利逻辑","回测证据","使用入口"],"surfaceModel":"浅色纸面行情单：主视觉是不断变化的波形线条，其余内容呈现在浅卡与细网格底色上，模拟盯盘面板。","palette":"暖纸白底 + 墨绿趋势线 + 玫红套利峰 + 蓝灰耦合副线。","typography":"大号数据感加粗数字用于价差异常值，正文用清晰人文无衬线，信号标签用紧凑大写。","desktopComposition":"横向主波形贯穿全宽作为第一信息层，其下以并列的耦合对（BTC↔Nasdaq、mNAV）卡片承载方法。","mobileTransformation":"波形缩为窄行顶通条，卡片流改为纵向时序列表，逐条展开耦合与回测细节。","mediaStrategy":"无本地素材，删除媒体区；波形、耦合对与价差标注纯排版实现。","forbiddenPatterns":["三卡指标墙","终端外壳","伪图表截图","功能卡片墙"],"compositionFingerprint":"连续价差波形作为贯穿视觉主线","evidenceBasis":"仓库事实：mNAV 套利、BTC-Nasdaq 耦合、自适应动量、回测"},{"id":"dir-gamma-inquiry-records","name":"研究笔记证物墙","projectMetaphor":"把仓库当作一本可持续增长的研究笔记本：每一条量化笔记（策略、指南、分类器、安全审计）像贴在墙上的索引卡片，按证据强度钉在不同深度。","signatureVisual":"逐层递进的证物卡墙：每组笔记以不同深度、不同标注颜色的索引卡呈现，卡边带日期与来源行，像研究者案桌上的钉板。","heroGeometry":"多张错位叠放的索引卡从视口右下角向左上堆叠，最上层卡片以一枚研究标签作为视觉起点。","sectionSequence":["研究笔记本首页","策略笔记","工具与审计笔记","LLM 指南","整理方法"],"surfaceModel":"浅暖稿纸底 + 深色索引标题条 + 彩色卡签，卡与卡之间留出手写感的空隙，模拟物理钉板层次。","palette":"稿纸米白 + 深墨标题 + 珊瑚、藏蓝、橄榄三色索引签 + 灰笔记正文。","typography":"手写感标题用于卡片标签，正文用可长读的无衬线，编号与来源用小号等宽。","desktopComposition":"墙面式错落网格，卡片按主题深度分前排/中排/后排三档进深感画面，来源行保持稳定对齐。","mobileTransformation":"卡片墙改为按主题折叠的纵向笔记本目录，逐项翻开单张索引卡。","mediaStrategy":"无本地素材，删除媒体区；用卡片堆叠、色彩索引签与进深排版代替。","forbiddenPatterns":["营销 Hero","功能卡片墙（无进深的等距卡阵）","浏览器外壳","发光网格粒子"],"compositionFingerprint":"分深度错落的证物索引卡墙","evidenceBasis":"仓库事实：投资策略、指南、ARDS-X 分类器、LAON VaultGuard、CASSANDRA AI、Awesome 系列"}],"selectedDirectionId":"dir-alpha-committee","selectionRationale":"仓库最核心的机制是多智能体与多 LLM 交叉核对并收敛成投资判断（multi-agent backtesting、Alpha Arena、多 LLM 交叉验证），委员会表决塔这一构图能把这一真实机制具象为证据流汇聚成置信度令牌，而非通用营销 Hero。它比价差波形更能覆盖整套内容（回测、核验、工具、指南），也比静态证物墙更能表达过程性与判断收敛，且没有本地素材时仍能靠排除排版自足。"},"designBrief":{"schemaVersion":"dodot-design-brief-v1","selectedDirectionId":"dir-alpha-committee","projectMetaphor":"多个 LLM 智能体组成一个投资委员会，分别核验论据并表决，最终收敛成一个带置信度的投资判断。","signatureVisual":"横向一排异形委员会席位（不同形状的节点代表不同 LLM），每个席位带一条证据链入线，汇聚到中央一枚表决令牌；令牌显示置信度百分比与\"通过/否决\"。","compositionFingerprint":"多源证据线汇入中央表决令牌","heroGeometry":"一组发散的证据链以不同角度汇入画面中心的一个菱形表决节点，节点四周标注当前议题、置信度与对冲符号。","sectionSequence":["表决现场","委员会席位","核验方法","反对票与对冲","研究库入口"],"surfaceModel":"深色会议厅舞台：每张席位是一块不规则的\"芯片茶几\"面板，表面带细密的证据纹线；交汇处令牌用高对比金属高光。","palette":"暗墨青底 + 琥珀表决色 + 冷白证据纹 + 单一玫红作为否决警示。","typography":"等宽衬线用于核对文字（象索引证编号），粗体无衬线用于议题标题，斜体小字用于唱票过程。","desktopComposition":"左右纵深式舞台：左列是席位与证据进线，右侧悬浮表决令牌与理由清单构成主视觉。","mobileTransformation":"席位列折叠为可顺序滑动的自定义 panel 组，令牌转为顶部全宽决策条。","mediaStrategy":"无本地截图素材，删除媒体区；用委员会席位+证据链线+令牌排版承担全部视觉。","evidencePlan":["以\"多智能体\"描述 NASDAQ/S&P500/加密市场 LLM 量化工具（description.facts 与 summary 证据），仅陈述工具范围不杜撰数字。","以\"回测、策略、核验、对冲\"对应仓库里的量化策略类条目（ARDS-X、Toss、mNAV 套利、Alpha Arena），只把存在的内容归入委员会席位。","以\"LLM 指南与安全审计\"对应仓库的 README 指南性与工具条目（VaultGuard、CASSANDRA AI、Awesome 系列），把它们标为研究库中的核验素材。"],"localeContract":{"zh-CN":"以\"AI 驱动投资研究与量化工具集\"为基调，用委员会表决隐喻组织章节，正文密度保持中文自然节奏。","en":"Ground in \"AI-powered investment research and quant tools\", organize sections with the committee-vote metaphor, keep English copy natural and concise."},"forbiddenPatterns":["营销 Hero","三卡指标墙","浏览器/终端/仪表盘外壳","发光渐变粒子场","伪图表/伪产品截图"],"reactBits":{"used":false},"antiTemplateCheck":{"recentCompositionFingerprints":[],"recentSignatureVisuals":[],"recentSectionSequences":[]}},"visualReview":{"schemaVersion":"dodot-automated-design-review-v1","status":"approved","summary":"zh-CN 与 en 两个页面忠实实现了项目专属 Brief（dir-alpha-committee · LLM 投资委员会表决）。选定方向的 signatureVisual（异形委员会席位+证据线汇入中央表决令牌）在 .vote-scene 中真实存在且与 brief 一致：6 个不同形状席位节点（circle/rhombus/pill）、6 条 SVG 证据线汇聚中央、菱形金属高光令牌显示 76% 与通过/PASS 及对冲注记。五大章节顺序（表决现场/委员会席位/核验方法/反对票与对冲/研究库入口）与 brief 的 sectionSequence 完全吻合。所有策略/工具事实（ARDS-X、LAON VaultGuard、CASSANDRA AI、自适应动量、mNAV 套利、BTC-Nasdaq 耦合、Alpha Arena、FRED/yfinance/DART 数据源）均可从 material.json 的 description/summary 证据逐条对应；边界声明（研究非建议、韩文为主、需 API 密钥）来自 summary.localizations.caveats。手机 CSS 形成结构重构（token 转为全宽静态决策块、席位节点转为 inline 词条、svgstream 隐藏、布局改单列），qa-report 显示移动端 scrollWidth=390=clientWidth 无溢出、fakeControlCount=0。双语章节逐段等价且英文自然（en cjkCount=0 纯英文）。未回退到营销 Hero/三卡指标墙/浏览器仪表盘外壳/发光粒子宫，卡片与网格均有 brief 证据用途。无假控件（fakeControlCount=0；vote-scene 标注 role=img 明确为装饰性场景，CTA/GitHub 为真实链接）。注：本环境无浏览器截图文件，审查仅依据源码与 qa-report 的 DOM/布局指标，未声称查看截图。","checks":{"projectFit":{"passed":true,"evidence":"选定方向 dir-alpha-committee 的核心隐喻'多智能体委员会表决、交叉核验、收敛为带置信度的判断'贯穿全局；页面语言、章节名（表决现场/核验方法/反对票与对冲/研究库入口）与 brief 的 sectionSequence 一一对应，未替换成通用主题。"},"evidenceBound":{"passed":true,"evidence":"仓库条目逐条有据：ARDS-X（FRED/yfinance）、LAON VaultGuard（Git 云私钥/离线）、CASSANDRA AI（KOSDAQ/DART/多 LLM 交叉验证）、自适应动量（Toss 模拟模式）、mNAV 套利与 BTC-Nasdaq 耦合与 Alpha Arena（来自 repository.description）、HTML 主语言、三条 caveats 均对应 material.json description/summary 的 claim 与 localization.caveats；未杜撰数字型仓库事实。"},"signatureVisual":{"passed":true,"evidence":".vote-scene 中 6 个异形席位节点（.seat-x 圆形、.seat-tria 菱形、.seat-h 不规则圆角）各带证据 ID（E-01…E-09），.svgstream 六条证据线自左右以不同角度汇聚至中心 (500,150)，中央 .token 菱形金属高光令牌显示 76% 与'通过/PASS'、对冲注记，完全复现 brief 的'多源证据线汇入中央表决令牌'指纹。"},"mobileTransformation":{"passed":true,"evidence":"max-width:560px 下 token 变 position:static 全宽决策块、席位节点转为 inline-flex 词条、svgstream 隐藏、.row/.seats/.verify-grid/.hedge-wrap 全部改单列——为结构性重构而非简单缩放；qa-report 显示 390px 视口 scrollWidth=390=clientWidth 无横向溢出、scrollHeight 与桌面不同。"},"localeParity":{"passed":true,"evidence":"两页结构逐节镜像（同 5 section、同节点/席位/verify 卡/caveats/L 链接），英文为自然地道改写而非逐字直译；qa-report 显示 zh textLength 1991/1996、en textLength 4630/4635，en cjkCount=0 纯英文；章节等价。"},"antiTemplate":{"passed":true,"evidence":"无营销 Hero/三卡指标墙/浏览器·终端·仪表盘外壳/无依据设备壳/发光渐变故子——验证卡为角色化核验方法卡（来源/可复现/多智能体交叉）而非指标墙；.seam 细密网格是 brief surfaceModel 指定的'证据纹线'而非单纯装饰；背景仅为暗色环境光洗底非粒子场。"},"noFakeUi":{"passed":true,"evidence":"qa-report fakeControlCount=0、emptyLinkCount=0；vote-scene 标 role=img aria-hidden 明确为装饰性示意，card 上的'表决·通过'均为纯文本状态标签；唯一的可交互 CTA/'进入研究库' 与 GitHub 链接均指向真实外部地址（vibequant.cc、github.com/gameworkerkim/vibe-investing）。"}},"issues":[{"code":"mobile-transformation-literal","severity":"warning","locale":"all","description":"brief 的 mobileTransformation 描述为'令牌转为顶部全宽决策条'与'席位列折叠为可顺序滑动的自定义 panel 组'；实际 CSS 把令牌做成视口中央的全宽静态决策块（position:static、居中），并把席位节点排成 inline 词条流，未实现'可顺序滑动 panel 组'的手势交互。此为描述性出入，不构成回退（仍是结构重构、无溢出、无假控件），故非 blocking。","repairInstruction":"如需更贴合 brief：可在移动端把令牌优先置于场景顶部形成全宽决策条，并考虑把席位节点封装为横向可横向滚动的顺滑 panel（需在 doc 内声明为装饰示意），保持 fakeControlCount=0 即可。"},{"code":"token-confidence-cosmetic","severity":"warning","locale":"all","description":"中央令牌的'76%'为构图性示意数字（meta 里也明示'收敛置信度'），而非 material.json 中的实际数据，已有'数值来自证据一致度，不是营销口号'的自我声明与 role=img 装饰语义，不构成杜撰仓库事实，但读者可能误读为真实指标。","repairInstruction":"可选：在令牌附近以更明确文字标注其为'构图示意'（如'示意收敛度 · 非实测值'），或在语义层面用 ARIA 说明为示意性图示，进一步避免被当成真实回测结果。两个 warning 均非阻塞。"}]},"repairEvidence":[],"provenance":{"executor":"opencode","model":"deepseek/deepseek-v4-flash","opencodeVersion":"1.1.58","promptVersion":"dodot-showcase-deepseek-v4","inputHash":"fe7bbb3b2fa0d93e9ba17c0a3ceba535eee8b238205101bf0024a7081e4caef7","artifactHash":"4d845ab5fb07781e8551b5c57808a349a075720931b83e9f71ac433f333a0d33","designSkill":"dodot-showcase-design"},"qaEvidence":{"zh-CN":[{"viewport":"desktop","passed":true,"localePassed":true,"metrics":{"language":"zh-CN","h1Count":1,"h2Count":5,"sectionCount":5,"textLength":1991,"cjkCount":1157,"latinLetterCount":327,"scrollWidth":1280,"clientWidth":1280,"scrollHeight":4878,"h1Top":142.796875,"bottomInset":213.625,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"5e3e2d05ff8137f3795b2da490f8a4ab9abbdad9fd6a369dc9045e89e293940f","middle":"11b5e8577cd7bdf98b342b832b3e39f84ada240626a1ad603ab60510d1b1e341","bottom":"da84afd5e482990b65cab5d2dc6e2e53286711ed55ede579b351979ab4bb26d8"},"consoleErrors":[],"networkRequests":[]},{"viewport":"mobile","passed":true,"localePassed":true,"metrics":{"language":"zh-CN","h1Count":1,"h2Count":5,"sectionCount":5,"textLength":1996,"cjkCount":1157,"latinLetterCount":327,"scrollWidth":390,"clientWidth":390,"scrollHeight":6148,"h1Top":134.796875,"bottomInset":234.921875,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"1c4ad34ca362c44298469f85e52b5c7146c7693ee733af23272d059ab8a07fc9","middle":"dc9a71fe487b6186cc1ed99b7947c2331938116ac2a5fa0f072130a151e6bc01","bottom":"bbfcb3369496d378e32dc7b0ddbe8700d5647bd3a6104ef16865ebbd7779423c"},"consoleErrors":[],"networkRequests":[]}],"en":[{"viewport":"desktop","passed":true,"localePassed":true,"metrics":{"language":"en","h1Count":1,"h2Count":5,"sectionCount":5,"textLength":4630,"cjkCount":0,"latinLetterCount":3604,"scrollWidth":1280,"clientWidth":1280,"scrollHeight":5076,"h1Top":142.796875,"bottomInset":213.46875,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"6e3143f32699e8a08f5d917296076dd5be7507f4bc83abb60bcb7baaddf1b481","middle":"9e3a6790f143ed0c0fd2dd6b023724bb55f1a2c2f0cffc9f45539a7d4063f5f8","bottom":"5b3f8031d9750e72d30647cbc17b86716dabe274c024d752b3df568c4a28e0b7"},"consoleErrors":[],"networkRequests":[]},{"viewport":"mobile","passed":true,"localePassed":true,"metrics":{"language":"en","h1Count":1,"h2Count":5,"sectionCount":5,"textLength":4635,"cjkCount":0,"latinLetterCount":3604,"scrollWidth":390,"clientWidth":390,"scrollHeight":6855,"h1Top":155.59375,"bottomInset":234.75,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"fc72e879acbf8d4ec14623b7d8dadeef9f989c55c95b7d1d41a3af0422877a31","middle":"3cfae146ec3a2cc80745a10c9aedf290c60193d4d9e40a6c168969e45cd3b8dd","bottom":"ab399783c6c8981c706257af459a8baba92c3975f29bb4495f294501cbffbc6a"},"consoleErrors":[],"networkRequests":[]}]}},"generatorVersion":"opencode-1.1.58:deepseek/deepseek-v4-flash:dodot-showcase-deepseek-v4","publishedAt":"2026-08-17T15:28:24.369Z"},"rawExtraction":{"projectRevisionId":"cmt2zh0b300nu6271bgfcr68j","projectVersion":1,"sources":[{"canonicalKey":"github:gameworkerkim/vibe-investing","role":"primary","revisionVersion":1,"url":"https://github.com/gameworkerkim/vibe-investing"}],"facts":[{"key":"repository_owner","value":"gameworkerkim","evidence":[{"path":"repository.owner","excerpt":"gameworkerkim"}]},{"key":"repository_name","value":"vibe-investing","evidence":[{"path":"repository.name","excerpt":"vibe-investing"}]},{"key":"description","value":"AI-powered Vibe Investing for NASDAQ, S&P500 & crypto: LLM quant trading tools, multi-agent backtesting, and data-driven market columns (mNAV arbitrage, BTC-Nasdaq coupling, Alpha Arena). 미국 주식·가상화폐 AI 투자 큐레이션·칼럼·트레이딩 봇.","evidence":[{"path":"repository.description","excerpt":"AI-powered Vibe Investing for NASDAQ, S&P500 & crypto: LLM quant trading tools, multi-agent backtesting, and data-driven market columns (mNAV arbitrage, BTC-Nasdaq coupling, Alpha Arena). 미국 주식·가상화폐 AI 투자 큐레이션·칼럼·트레이딩 봇."}]},{"key":"primary_language","value":"HTML","evidence":[{"path":"repository.language","excerpt":"HTML"}]},{"key":"topics","value":"agentic-ai, ai-search, algorithmic-trading-dotnet, backtesting, bitcoin, claude, cryptocurrency, cryptocurrency-analyser, defi, earnings-momentum, earnings-surprise, financial-llm, llm, multi-agent, quantitative-trading, sentiment-analysis, stock-market, stock-screening, trading-bot","evidence":[{"path":"repository.topics","excerpt":"agentic-ai, ai-search, algorithmic-trading-dotnet, backtesting, bitcoin, claude, cryptocurrency, cryptocurrency-analyser, defi, earnings-momentum, earnings-surprise, financial-llm, llm, multi-agent, quantitative-trading, sentiment-analysis, stock-market, stock-screening, trading-bot"}]},{"key":"source_count","value":"1","evidence":[{"path":"projectRevision.sources","excerpt":"github:gameworkerkim/vibe-investing"}]}],"taxonomy":{"category":"development_tools","subcategory":"data_ai_tools","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"AI-powered Vibe Investing for NASDAQ, S&P500 & crypto: LLM quant trading tools, multi-agent backtesting, and data-driven market columns (mNAV arbitrage, BTC-Nasdaq coupling, Alpha Arena). 미국 주식·가상화폐 AI 투자 큐레이션·칼럼·트레이딩 봇."}]},"lifecycle":{"stage":"active","confidence":0.75,"evidence":[{"path":"repository.pushedAt","excerpt":"2026-08-17T06:02:44.000Z"}]},"userIntents":[{"slug":"build_software","label":"构建或改进软件","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"AI-powered Vibe Investing for NASDAQ, S&P500 & crypto: LLM quant trading tools, multi-agent backtesting, and data-driven market columns (mNAV arbitrage, BTC-Nasdaq coupling, Alpha Arena). 미국 주식·가상화폐 AI 투자 큐레이션·칼럼·트레이딩 봇."}]},{"slug":"automate_work","label":"自动化工作流程","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"AI-powered Vibe Investing for NASDAQ, S&P500 & crypto: LLM quant trading tools, multi-agent backtesting, and data-driven market columns (mNAV arbitrage, BTC-Nasdaq coupling, Alpha Arena). 미국 주식·가상화폐 AI 투자 큐레이션·칼럼·트레이딩 봇."}]},{"slug":"analyze_data","label":"分析与理解数据","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"AI-powered Vibe Investing for NASDAQ, S&P500 & crypto: LLM quant trading tools, multi-agent backtesting, and data-driven market columns (mNAV arbitrage, BTC-Nasdaq coupling, Alpha Arena). 미국 주식·가상화폐 AI 투자 큐레이션·칼럼·트레이딩 봇."}]},{"slug":"learn_explore","label":"学习与探索","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"AI-powered Vibe Investing for NASDAQ, S&P500 & crypto: LLM quant trading tools, multi-agent backtesting, and data-driven market columns (mNAV arbitrage, BTC-Nasdaq coupling, Alpha Arena). 미국 주식·가상화폐 AI 투자 큐레이션·칼럼·트레이딩 봇."}]}],"readme":{"headings":["vibe investing","NEW · 당신의 LLM에 넣는 그 프롬프트, 검증된 건가요? — Awesome LLM Prompts for Finance & Quant","NEW · 휴먼을 갈아 넣은 텐센트 Hunyuan Hy3 소개 - Hy3 Getting Started 가이드 aka 토큰 거지를 위한 텐센트의 대답","NEW · LLM Wiki 3종 비교 — DeepWiki · Google Code Wiki · OpenWiki","NEW · Claude Fable 5 전체 시스템 프롬프트 유출 — LLM 개발자를 위한 인사이트 모음","NEW · [꿀팁 공개] 토큰 거지를 위한 토큰 최적화 유틸리티 caveman + rtk — 최대 95% 절감","NEW · 나와 팀을 위한 지식 검색 LLM RAG — Quivr: 오픈소스 Second Brain","NEW · GitHub 오픈소스 트레이딩 봇 17종 비교 분석 v2.0 — AI/LLM 확장판 (6월 10일 update)","NEW · 이번에 터진 콘텐츠 — Toss x AMQS 퀀트 대시보드","NEW · 하락을 판단하는 퀀트 — ARDS-X 레짐 분류기 (소스 + 대시보드)","NEW · 시크릿은 새기 전에 — LAON VaultGuard (npm 패키지 + VS Code 확장)","NEW · 코스닥 시장의 카산드라 — CASSANDRA AI (LLM 기반 한계기업 공시 분석)","가장 반응이 좋았던 투자 칼럼 3선 (LinkedIn 90일 실측)","이 레포가 향하는 곳 — 멀티-LLM 투자위원회","Vibe Investing이란?","내가 만든 Awesome 시리즈","다른 Awesome 시리즈와의 차이","장르별 큐레이션","전략 (Trading Strategy)","칼럼 (Investment Idea Column)"],"commands":[]},"changeSet":[{"field":"repository","beforeHash":"cc6ee842d66ffb593a982c2ad4fac1c2f9562491b60a11e2eae9686d9ea0b109","afterHash":"7694ea8720b2d000bb4a9ab69a0e6cd7b539cc11f06f22db230f8e87a85ad740"}]},"publishedAt":"2026-08-16T12:08:30.767Z","updatedAt":"2026-08-21T13:26:50.362Z","summary":{"id":"cmt2zh0b700nw6271geotf9fa","projectType":"other","oneLine":"AI投资研究、量化策略与工具集","overview":"这是一个AI驱动的投资研究资源库，涵盖美股、S&P500和加密货币。它提供LLM量化交易工具、多智能体回测、市场专栏，以及多个实用指南，如金融LLM提示词、腾讯Hunyuan Hy3入门、LLM Wiki比较、token优化工具等。还包含Toss量化仪表盘、ARDS-X衰退分类器、LAON VaultGuard安全审计和CASSANDRA AI等具体项目。适合对AI投资和量化交易感兴趣的开发者与投资者。","highlights":["多智能体回测工具","市场专栏与策略","实用LLM指南","开源量化项目"],"bestFor":["AI投资研究者","量化交易爱好者","LLM应用开发者"],"caveats":["内容以韩文为主，部分指南为韩文","项目多为研究性质，非投资建议","部分工具需要API密钥或外部数据"],"categorySuggestion":"other","subcategorySuggestion":"uncategorized","lifecycleStages":["idea_requirements","testing_quality"],"audienceIntents":["discover","learn","evaluate"],"confidence":0.7,"evidence":[{"claim":"该仓库是一个AI驱动的投资研究资源库，涵盖美股、S&P500和加密货币，提供LLM量化交易工具、多智能体回测和市场专栏。","sourceType":"description","locator":"https://github.com/gameworkerkim/vibe-investing"},{"claim":"README包含多个指南，如金融LLM提示词、腾讯Hunyuan Hy3入门、LLM Wiki比较、Claude Fable 5系统提示分析、token优化工具caveman+rtk、Quivr RAG指南等。","sourceType":"readme","locator":"https://github.com/gameworkerkim/vibe-investing"},{"claim":"仓库包含Toss量化仪表盘，将自适应动量量化策略应用于韩国股票和ETF，提供模拟模式，无需API密钥即可运行。","sourceType":"readme","locator":"https://github.com/gameworkerkim/vibe-investing/tree/main/Toss"},{"claim":"仓库包含ARDS-X衰退分类器，使用FRED和yfinance的真实数据自动分类市场下跌类型，并提供可视化仪表盘。","sourceType":"readme","locator":"https://github.com/gameworkerkim/vibe-investing/tree/main/01.Trading%20Strategy/ARDS%20%E2%80%94%20Adaptive%20Recession-Defensive%20Strategy%20for%20AI_QQQ"},{"claim":"仓库包含LAON VaultGuard，一个跨平台的多LLM安全审计工具，用于检测Git仓库中的云私钥泄露，支持离线模式。","sourceType":"readme","locator":"https://github.com/gameworkerkim/vibe-investing/tree/main/LAON_VaultGuard"},{"claim":"仓库包含CASSANDRA AI，一个基于LLM的韩国KOSDAQ市场可疑公司披露分析系统，使用DART数据并支持多LLM交叉验证。","sourceType":"readme","locator":"https://github.com/gameworkerkim/vibe-investing/tree/main/CASSANDRA%20AI"}],"publishedAt":"2026-08-17T13:24:08.018Z"},"github":{"url":"https://github.com/gameworkerkim/vibe-investing","ownerAvatarUrl":"https://avatars.githubusercontent.com/u/5234949?v=4","starsCount":323,"forksCount":98,"language":"HTML","syncedAt":"2026-08-17T13:20:52.563Z","license":null,"defaultBranch":"main","archived":false,"disabled":false,"lastChangedAt":"2026-08-17T06:02:44.000Z","missing":false},"tags":[],"updates":[]},"entryPoint":"discovery","claim":{"state":"unclaimed","verified":false},"initialView":"beautified","community":{"viewer":null,"state":{"likesCount":0,"creativeCount":0,"followersCount":0,"commentsCount":0,"liked":false,"creative":false,"followed":false},"comments":{"comments":[],"nextCursor":null}}}]}],"$L24"]

不是一个人的盘感，而是一场多智能体的表决。

证据入线

交叉核验

结算置信度

每一类内容，都是会议桌上的一席。

自适应动量 · 策略席

ARDS-X · 衰退判定席

LAON VaultGuard · 纪律席

CASSANDRA AI · 警戒席

判断之前，先核验证据来源。

真实数据入线

模拟模式可跑

同一结论多重读出

容许反对票，才是成熟的投资委员会。

研究性质，非投资建议

内容以韩文为主

部分工具需要密钥

走进席位，开始你的核验。

vibequant.cc · 研究库

gameworkerkim/vibe-investing

入会须知 · 三条边界

Not one trader's hunch — a multi-agent vote.

Evidence enters the line

Cross-check

Settle on confidence

Each body of work fills a chair.

Adaptive Momentum · strategy

ARDS-X · downturn classification

LAON VaultGuard · discipline

CASSANDRA AI · caution

Before deciding, audit the source.

Real data in the line

Mock mode that runs

Same call, several reads

A grown-up committee welcomes the no vote.

Research, not advice

Mostly Korean content

Some tools need keys

Take your seat, start verifying.

vibequant.cc · research library

gameworkerkim/vibe-investing

Before you sit in · three boundaries