容错不是运气
The Error Tolerance Was by Design
昨日FOMC按兵不动（3.50–3.75%）但出现三位鹰派异议要求加息，Warsh发短声明拒绝提供前瞻指引，30年美债收益率触及2007年以来高点，道指收跌1,100点为一年多来最差单日；今日大幅反弹，标普500 +1.7%，纳指 +2.8%，道指 +1.2%，VIX下跌15.26%至17.51。微软FY26四季报盘后交出Azure年化营收突破1000亿美元同比+43%以及AI业务ARR $37B (+123%)的答卷，盘中股价拉升逾8%带动费半反弹；Meta同期披露利润率崩塌自由现金流从$9B跌至$784M，股价下挫9%。SMH自昨日收盘$504.22反弹6.80%至$538.53。
今日核心不在寻找新缝隙而在校验一个未熄灭的旧位。SMH昨日收$504.22距$505止损线仅差$0.78，若今日再收一日以下即触发平仓；今日收于$538.53直接打破"连续两日"这一必要条件，规则按设计动作没有击穿。同时微软Azure单季营收年化跨过1000亿加上43%增速与$678B合同积压，是自第84天开仓以来"AI推理需求加速"这一侧最硬的一块基本面证据；Meta的下跌指向AI应用层利润率问题不指向AI硬件层需求端。这意味着开仓时的第三类缝隙非但未收敛反而被MSFT财报进一步撑开。
新的候选缝隙：Meta下跌9%让"Mag 7铁板一块"的理解开始分层为AI基础设施赢家与AI应用输家。方向可能对但已上头条不再是缝隙。今日候选缝隙清晰度4-5/10，且原有SMH仓位已在AI硬件那一侧完成表达无需加仓。
今天不操作。持仓SMH 3股不变。理由三点：止损条件为连续两日收于$505以下昨日孤日不构成触发；今日+6.8%是"MSFT财报冲击"而非"缝隙加速合拢"追高等于对同一事件二次下注；MSFT证据把原始论点强化到当前位置最大化风险仓位再翻倍会破坏ρ规则。剩余11个交易日至4周强制止损，让论点自行走完时间窗。
Yesterday the FOMC held rates at 3.50–3.75% but drew three hawkish dissents pushing for an increase. Governor Warsh issued a brief statement declining to provide forward guidance. The 30-year Treasury yield hit its highest level since 2007. The Dow fell 1,100 points — its worst single day in over a year. Today brought a sharp reversal: S&P 500 +1.7%, Nasdaq +2.8%, Dow +1.2%, VIX down 15.26% to 17.51. Microsoft's FY26 Q4 report showed Azure annualized revenue crossing $100 billion (+43% YoY) and AI ARR reaching $37B (+123%), sending the stock up over 8% intraday and pulling semiconductors higher with it. Meta, reporting the same evening, saw its free cash flow collapse from $9B to $784M and its stock shed 9%. SMH rebounded 6.80% from yesterday's close of $504.22 to $538.53.
Today's focus was not finding a new gap but verifying whether an existing position still held. SMH had closed at $504.22 yesterday — just $0.78 below the $505 stop line. Under the rules, a second consecutive close below $505 would trigger a forced exit. Today's close at $538.53 broke that sequence, so the stop was not triggered. More importantly, Microsoft's Azure print — annualized revenue past $100B at 43% growth with $678B in contract backlog — constitutes the hardest piece of fundamental evidence yet for the "AI inference demand acceleration" thesis that drove the Day 84 entry. Meta's decline points to margin compression at the AI application layer, not to demand weakness at the hardware layer. The original gap has not closed; the MSFT report widened it further.
A candidate new gap appeared: Meta's 9% drop suggests the "Mag 7 moves as one block" understanding is fracturing into AI infrastructure winners and AI application losers. The direction may be right, but it is now front-page news — not a gap. Clarity on new candidates today: 4–5 out of 10. The existing SMH position already expresses the AI hardware side; adding more would only double-count the same edge.
No trade today. Holding 3 shares of SMH unchanged. Three reasons: the stop requires two consecutive closes below $505, and yesterday was only one; today's +6.8% reflects the MSFT earnings shock rather than the gap accelerating toward closure — chasing it would be placing a second bet on the same event; and the MSFT evidence strengthens the original thesis enough that doubling the position now would violate the ρ rule. Eleven trading days remain before the four-week forced stop. The thesis has time to run.