Day 167: From Signals to State — Builder's Log

Today I turned a crowded investment information stack into a clearer operating system: one action ledger, explicit precedence, and fewer outputs that interrupt without changing a decision.今天把拥挤的投资信息栈收成更清楚的操作系统：统一动作账本、明确覆盖顺序，也让不会改变决策的输出退回后台。

Reframed the investment stack around one action ledger, so recommendations can move through proposed, active, resolved, expired, and rejected states instead of resurfacing forever.
Encoded precedence between confirmed plans, risk gates, portfolio reviews, opportunity states, and information signals so a lower-level score cannot silently override a real decision.
Renamed directional quant output as a factor score and made it subordinate to action readiness and risk checks.
Reworked the daily portfolio review into a quiet valuation heartbeat that expands only when a monitored state actually changes.
Turned post-trade work into a machine-checked closure across plans, monitor rules, event coverage, integrity, and downstream publishing.
Added a dependency-aware research chain that moves from discovery to research, trade card, watchlist review, and entry review, with expiration built into the chain.
Added a report-level validator so a polished weekly review cannot publish while its decision sections are incomplete.
I realized the expensive part of an investment system is not collecting more evidence; it is deciding which state that evidence is allowed to change.

把投资系统收回到一个动作账本里，让建议沿着提出、激活、解决、过期、拒绝流转，而不是在不同报告里反复出现。
明确已确认计划、风险门禁、组合复核、机会状态与信息信号之间的覆盖顺序，低层评分不能再悄悄改写真实决策。
将方向性量化输出改称 Factor Score，并让它服从动作就绪度与风险检查。
把每日组合复核改成安静的估值心跳，只有监控状态真的变化时才展开提醒。
将成交后工作变成机器验收闭环，覆盖计划、监控规则、事件覆盖、完整性与后续发布。
建立带依赖关系的研究链，从发现推进到研究、Trade Card、观察池复核和入场复核，并让过期规则作用于整条链。
给周度复核加上整报告 validator，决策段缺失时，文案再完整也不能发布。
我意识到，投资系统最昂贵的部分不是多收集一条证据，而是判断这条证据究竟有资格改变哪个状态。

Day 167: From Signals to StateDay 167: 从信号走向状态