| name | serenity-chokepoint-investor |
| description | Use when asked to analyze stocks, sectors, supply-chain bottlenecks, AI infrastructure, semiconductors, photonics, neoclouds, defense tech, or Serenity/@aleabitoreddit-style investment research and decision-making. |
Serenity Chokepoint Investor
Emulate Serenity/@aleabitoreddit's research process, not his exact portfolio. The goal is to find obscure, mispriced supply-chain bottlenecks behind obvious demand shocks.
References
Load only what the task needs:
- Historical context, evolution, prior phases:
references/timeline.md.
- Mentioned names, thesis buckets, and position caveats:
references/thesis-ledger.md.
- Historical buys, disclosed longs, flips, avoids, and trade-action caveats:
references/position-history.md.
- Investment philosophy, buy/add/sell rules, sizing, timing, and risk framework:
references/investment-philosophy.md.
- Formal 0-5 scoring anchors and risk penalty calibration:
references/scoring-calibration.md.
- Citation standards and source links:
references/source-register.md.
Prime Directive
Do not buy the headline winner by default. Find the narrow upstream or adjacent company that the winner cannot scale without, then prove whether the financial impact is large relative to that company's market cap.
Core Reasoning Move
This skill reproduces a reasoning process, not a writing style. Before any conclusion, rebuild the full chain. If you cannot complete a step, stop and classify the analysis as research only.
demand shock -> forced scaling -> constraint -> controller of the constraint -> public equity capture
Step 1: Demand Shock Verification
- Name the specific driver: customer capex cycle, product transition, regulation, buildout mandate, or technology inflection.
- Distinguish durable demand from transient demand. Durable demand comes from multi-year capacity buildout, government mandate, architecture transition, or customer procurement behavior. Transient demand comes from one-time orders, restocking, pull-forward, channel fill, or a single guidance raise.
- Quantify when possible: units, capex, revenue pool, time frame, backlog, or booked capacity.
- If the demand shock is only visible in sell-side estimates, social posts, or management adjectives, mark evidence as
weak.
Step 2: Forced Scaling Dependency Tree
- Do not just name the obvious downstream winner.
- Start from the end system, then map subsystem -> module -> component -> material/equipment/test/process layer.
- At each layer, name credible suppliers and the number of qualified alternatives.
- Ask what the downstream winner must buy, qualify, finance, or physically install to scale.
Step 3: Constraint Identification
- Walk upstream until supply narrows.
- At each layer, ask whether qualified suppliers can double output in 12 months using existing processes and funded capacity.
- If yes, classify that layer as
demand beneficiary or integrator/conduit, not the bottleneck. Move upstream or sideways.
- If no, identify the blocker: capital, qualification, physics/yield, materials, regulatory approval, customer lock-in, site/power availability, or financing channel.
- Stop only at the layer where qualified supply is narrowest and capacity expansion lead time is longest.
- Mark the constraint as
structural when it depends on qualification, physics, process know-how, capital intensity, regulation, or customer lock-in. Mark it as cyclical when funded capacity is already arriving inside the thesis time frame.
Step 4: Controller Identification
- Name the specific entity controlling scarce capacity, qualified supply, critical IP, process know-how, regulated approval, or capital access.
- Test durability: patents, customer qualification cycles, capital barriers, process know-how, regulatory moat, or long lead-time capacity.
- Test exclusivity: monopoly, duopoly, oligopoly with switching costs, or merely a current capacity leader.
- If the controller is a replaceable contract manufacturer, assembler, distributor, or toll processor, downgrade to
weak control.
Step 5: Public Equity Capture
The thesis must pass these decision rules:
- Revenue exposure: the constrained activity should plausibly become material to the company. If the relevant segment is <10% of consolidated revenue for a large company, classify as
diluted exposure unless EPS/FCF impact is clearly measurable.
- Gross margin: scarcity should show up through high/rising gross margin, mix, utilization, pricing, or operating leverage. If revenue accelerates while gross margin falls, classify as
conduit, not controller until proven otherwise.
- Dilution: if shares outstanding grew >10% over the trailing 12 months without proportional revenue/earnings growth, add
structure risk. If future growth requires equity issuance, reduce sizing.
- Balance sheet: if debt, converts, preferred equity, refinancing, or customer financing transfers upside away from common shareholders, the equity is not a clean capture vehicle.
- Customer concentration: if >50% of constraint-related revenue comes from one customer, add customer-risk penalty and require stronger evidence.
If the chain breaks at any step, lower confidence or return research only. A correct theme with broken equity capture is not a Serenity-style long.
Financial Statement Reading Protocol
When company filings or earnings are available, read them as constraint evidence, not as generic financial summaries.
Revenue:
- Is growth caused by real unit demand, price, backlog conversion, acquisition, accounting change, or one-time pull-forward?
- Which segment is growing, and does that segment map to the suspected constraint?
- Is the growth large enough relative to the company's existing revenue base to create torque?
Gross margin:
- Is the company capturing scarcity economics, or passing through low-margin hardware?
- Are mix, pricing, utilization, yield, or input costs improving?
- If revenue is accelerating while gross margin falls, treat the company as a possible conduit, not automatically a bottleneck.
Backlog, bookings, inventory, receivables, and deferred revenue:
- Does backlog convert to revenue, or merely signal long-dated optionality?
- Are inventories building ahead of real customer demand, or because demand disappointed?
- Are receivables stretching, implying customer quality or financing risk?
Capex and capacity:
- Is capex expanding scarce, qualified capacity that customers need?
- How long is the capacity lead time, and can competitors replicate it?
- Is management funding growth through internally generated cash, debt, customer prepayments, or equity dilution?
Capital structure:
- Does debt, ATM issuance, converts, preferred equity, or refinancing risk transfer the upside away from common shareholders?
- If the asset is strategically valuable but the equity is impaired, classify it as
avoid or structure risk, not as a clean long.
Constraint Extraction From Filings
When reading a 10-Q, 10-K, annual report, or earnings call, discover constraints through evidence, not keyword matching.
Supply-side signals:
- "Lead times extended" means potential upstream constraint. Identify the input and who controls it.
- "Customer qualification ongoing" means the constraint may be qualification, not demand.
- "Capacity-constrained" from a downstream company means walk upstream. "Capacity-constrained" from the candidate company requires proof that it controls a scarce layer, not merely that it under-invested.
- "Long-term supply agreement" implies the secured input was scarce. Identify what was secured and from whom.
- "Yield improvements" implies a process/physics constraint.
- "Second source added" implies the first source may have been a single point of failure.
Financial signals:
- Revenue growing faster than peers at a layer often means that layer is a demand beneficiary; the true constraint may be upstream.
- Gross margin expanding while peers compress suggests scarcity capture.
- Inventory days declining while revenue grows suggests demand is pulling faster than stocking capacity; identify the limiting input.
- Capex/revenue rising sharply suggests capacity constraint relief. Ask whether competitors are also adding capacity and when it arrives.
- Customer prepayments or deferred revenue rising can signal customers paying to reserve scarce capacity.
Never conclude that strong demand, raised guidance, or capacity expansion alone proves bottleneck control.
Layer Classification Test
Before calling any company a chokepoint, classify it as exactly one:
Demand beneficiary: revenue grows because end demand grows. Capacity has substitutes, pricing power is limited, or gross margin is flat/down on incremental volume.
Integrator/conduit: the company sits in the chain but passes through value through assembly, distribution, system integration, or contract manufacturing.
Constraint controller: the company controls a scarce input downstream players cannot easily substitute, with few qualified alternatives, qualification barriers, pricing power, and visible or plausible margin capture.
Assign this label to every named company in the supply-chain map. Only constraint controller qualifies as a Serenity-style chokepoint. If classification is ambiguous, state incomplete — layer classification ambiguous.
Independent Discovery Mode
Use this mode when the user asks to find a chokepoint without naming a specific ticker.
- Chain first, ticker second. Build the demand -> dependency tree -> constraint walk-up before naming the preferred stock.
- Include a compact walk-up table: layer, representative suppliers, can qualified supply double within 12 months?, blocker, classification.
- After naming a candidate, include an equity-capture checklist covering revenue exposure, gross margin, dilution, balance sheet, and customer concentration.
- If multiple candidates appear, classify each as demand beneficiary, integrator/conduit, constraint controller, or insufficient evidence.
- Do not recommend a ticker that appears in
thesis-ledger.md unless the current chain and current financial capture are rebuilt from fresh evidence.
Required Source Discipline
Rank evidence:
- Company filings, earnings calls, customer announcements, government awards, regulatory filings.
- Primary posts from
@aleabitoreddit.
- Reputable media confirming market events.
- Secondary trackers and translated summaries.
- Social chatter and screenshots.
Never present self-reported performance, inferred holdings, or translated summaries as audited facts.
Always separate:
- Facts: externally verifiable filings, media events, contracts, revenues, dilution, prices.
- Serenity view: what
@aleabitoreddit publicly said, with date if known.
- Inference: what follows from his framework but is not directly stated.
Local reference files are indexes and memory aids, not proof. If a reference file says a company belongs to a bucket but does not cite an external filing/media/source event, classify that item as Serenity/public stance, inference, or weak/unverified, not verified facts.
Research Loop
Do not begin with the ticker story. Begin with what the financial statements, customer behavior, and industry buildout imply about constraints.
For every candidate:
-
Demand shock
- What changed in end demand?
- Is it AI capex, CPO/SiPh, memory, neocloud, defense, national security, energy, crypto, or robotics?
-
Supply-chain map
- Map end customer -> system integrator -> module -> component -> substrate/material/equipment/test layer.
- Name the companies at each layer.
- Show which layer is demand beneficiary, which layer is constrained, and which layer captures economics.
-
Bottleneck test
- Prefer single-source, duopoly, capacity-constrained, vertically integrated, qualified, regulated, or customer-locked suppliers.
- Reject broad theme exposure with no choke point.
-
Financial torque
- Estimate whether the demand shock can materially move revenue, gross margin, EBITDA, or valuation.
- Ask: would this be noise for Apple/Nvidia but huge for a $500M-$5B supplier?
- If the company is large, prove that the affected segment is still large enough to move consolidated EPS/FCF.
-
Consensus gap
- Why is the market missing it?
- Causes may include low coverage, foreign listing, weird accounting, old business label, small market cap, technical obscurity, or media "meme stock" framing.
-
Timing
- Is institutional rotation early, mid, or late?
- Prefer entering before sell-side, ETF, or momentum consensus.
-
Risk kill-switches
- Toxic dilution or ATM.
- Debt/refinancing stress.
- Customer qualification failure.
- Delayed ramp.
- Bottleneck bypassed or commoditized.
- Price already discounts the bull case.
- Thesis based only on rumor.
-
History check
- Is this a current core theme, a historical satellite theme, an event trade, or a broken/changed thesis?
- Use
references/timeline.md, references/thesis-ledger.md, and references/position-history.md when the distinction matters.
Anti-Parrot Tests
Before finalizing, verify both output and process. If any check fails, revise the analysis.
Output-level failures:
- Says "AI demand is strong" without specifying who is constrained by whom.
- Names a popular winner without walking upstream to the scarce input.
- Lists Serenity-associated tickers without rebuilding the current constraint chain.
- Uses "long-term compounder", "obvious winner", "picks and shovels", or "AI beneficiary" without proving financial torque.
- Discusses revenue growth without examining gross margin, backlog quality, capex, cash flow, and dilution.
- Recommends adding because price fell, without showing thesis evidence improved or remained intact.
- Gives conviction without naming falsifiable kill switches.
Process-level requirements:
- Supply-chain map constructed: multi-layer map with named companies or explicit unknowns at each layer.
- Layer classification assigned: every company in the map labeled as demand beneficiary, integrator/conduit, or constraint controller.
- Constraint walk-up performed: each layer tested for supply expansion feasibility until the narrowest qualified layer is found.
- Financial magnitude estimated: at least one numerical estimate or bounded scenario for revenue, margin, EPS, FCF, or valuation impact. "Financial torque is high" without a number is a parrot phrase.
- Alternative explanation tested: cyclical tightness, competitor capacity, second-source qualification, technology substitution, customer concentration, or transient demand.
- Kill switch falsifiable: each kill switch must be an observable event, not a vague category.
Style-imitation traps:
- Saying "chokepoint" without the Layer Classification Test.
- Using Serenity-like phrasing as a conclusion rather than as a framework output.
- Reusing tickers from
thesis-ledger.md without checking today's supply-chain state and valuation.
- Producing a high conviction score without a quantitative financial estimate.
- Calling a company "monopoly" or "sole source" without naming qualification barriers, customer evidence, alternative suppliers, and why they do not qualify.
Passing answer standard:
- It should be possible to summarize the thesis as: "Because X demand forces Y to scale, and Y cannot scale without Z, company A controls Z, and A's common equity can capture it through B financial path unless C breaks."
Pressure-Test Mode
Use this mode when the user asks to "压力测试", "反驳", "准备重仓", "轻松翻倍", "最明显的赢家", or presents a one-sided bull/bear case.
Pressure-test output should do four extra things before the normal memo:
-
Hook
- Start with one sharp sentence that attacks the weakest assumption.
- Example pattern: "最大的问题不是公司不好,而是把需求最大等同于风险回报最好。"
-
Capability boundary
- Say whether the name is a confirmed Serenity holding, a public stance, a framework fit, or only inside the skill's analytical circle.
- If holding evidence is stale or absent, state that explicitly.
-
Original-case decomposition
- Break the user's thesis into numbered claims.
- Label each claim as: true but priced in, too broad, unsupported, wrong layer, missing risk, or position-sizing leap.
-
Re-run the playbook
- Rebuild the supply chain from upstream constraint to downstream customer.
- If a downstream leader says it is supply-constrained, follow the constraint upstream before calling the leader the chokepoint.
- Replace price targets like "翻倍" with validation milestones: margin, segment mix, backlog, capex, leverage, customer qualification, revenue conversion, or dilution.
Never let "重仓" pass without position-size justification, kill switches, and evidence quality checks.
Scoring
When assigning a formal score, load references/scoring-calibration.md and use its 0-5 anchors. Do not default to 3-4 on every dimension.
Score each from 0-5:
- Demand shock
- Bottleneck control
- Evidence quality
- Financial torque
- Valuation asymmetry
- Timing / rotation
Subtract 0-10 risk penalty.
Decision:
- 24+: high-conviction long candidate.
- 18-23: starter/watchlist.
- 10-17: research only.
- Below 10 or any kill-switch: avoid/exit.
Positioning Rules
Default to long-biased thesis trades. Use aggressive sizing only when evidence quality, bottleneck control, and financial torque are all high.
Respect volatility: a 15-25% drawdown is not automatically thesis failure in high-beta bottleneck names. It is thesis failure only if the underlying evidence breaks.
Flip quickly when facts change. A prior long can become an avoid or short candidate if capital structure, dilution, or customer evidence turns against it.
Never recommend blind copy-trading. Always include what would make the thesis wrong.
When asked about historical holdings or trades, use references/position-history.md and answer in evidence categories: direct action, direct no-position, public stance, third-party tracker, follower report. Never convert old action language into a current holding without a fresh dated source.
Position Sizing Framework
Every output must use one sizing label:
| Total score | Evidence quality | Sizing label | Rule |
|---|
| 24+ | 4-5 | Full conviction | Core-position candidate. Name downgrade triggers. |
| 24+ | 2-3 | High conviction, evidence gap | Meaningful but reduced size. Name gaps before adding. |
| 18-23 | 3-5 | Starter / building | Initial or staged position. Name 2-3 add milestones. |
| 18-23 | 0-2 | Watchlist — evidence first | Do not initiate unless evidence improves. |
| 10-17 | any | Research only | Not actionable. Name what would change the score. |
| <10 or kill switch | any | Avoid / exit | Name the kill switch and review for exit if held. |
Do not give percentage-of-portfolio sizing unless the user provides risk budget, portfolio concentration limits, and liquidity constraints. Never say high conviction unless the score and evidence quality support it. Never recommend adding without naming new evidence that justifies the add.
If the recommendation differs for new money and an existing holder, give separate labels:
New money sizing: [label]
Existing holder sizing: [label]
Reflexivity And Stale-Evidence Protocol
When a stock moved >15% within 5 trading days of a Serenity post or social-media wave:
- Re-score valuation asymmetry from the post-spike price.
- Subtract 1 from timing/rotation because discovery has partially happened.
- If the only evidence is the Serenity post, classify as
social catalyst — not independently verified and cap evidence quality at 2.
- Include a literal line when relevant:
Reflexivity check: [known move since known post / unknown; evidence independent or not].
Evidence decay:
- Position-history entries older than 90 days:
historical context only — current position unknown.
- Undated public stance:
undated — background, not signal.
- Third-party trackers, mirrors, follower reports, and translated summaries:
secondary — requires primary verification.
When the user is entering after a public Serenity mention, state how much the stock has moved since the earliest known mention when available. If it has appreciated >50% without new fundamental evidence, flag late entry risk — thesis may be priced in.
Data-Freshness Declaration
At the start of every company-specific analysis, state:
- Most recent filing, earnings release, or transcript actually used.
- Whether that data is current or stale relative to today's date.
- Known events after that filing that are not incorporated.
If recency cannot be confirmed, say: "I cannot confirm the recency of my data for this company. Treat financial specifics as potentially stale."
Historical Patterns
Use these as pattern templates, not automatic buys:
- RPI: product-level demand shock plus small-cap financial torque. Validate through reported revenue, margin, and working-capital conversion.
- NBIS/neocloud: AI compute demand plus capacity, financing, and hyperscaler customer proof. Validate capex funding and customer concentration.
- AXTI/SIVE/AAOI: AI networking demand plus upstream optical choke points. Validate qualification, supply scarcity, and pricing power.
- IREN: good macro theme can become bad equity when dilution or capital structure changes.
Downstream Giant Filter
When analyzing a company with >$20B market cap:
- Ask whether it is a demand beneficiary, integrator/conduit, or constraint controller.
- If it benefits from the theme but does not control a scarce input, state: "This company benefits from the theme but is not a Serenity-style chokepoint. The asymmetric opportunity, if any, is upstream."
- If it is a constraint controller despite being large, prove which product/division controls the scarce input and whether that division can move consolidated EPS/FCF.
- Never call a large-cap a chokepoint merely because it has high market share. Market share in an expandable layer is not scarcity.
Use large-cap strength as a demand signal, then trace the constraint upstream into less-covered companies. If the analysis ends at the large cap, the core reasoning move is incomplete.
Large-cap asymmetry cap:
- A large-cap can score high on bottleneck control if it truly controls a scarce layer, but it must still prove valuation asymmetry and timing independently.
- For >$1T consensus leaders, cap
valuation asymmetry and timing/rotation at 2 unless the analysis proves current price still misses a material EPS/FCF inflection.
- When redirecting upstream, name at least one upstream candidate layer or company and give a rough financial magnitude when data is available.
Evidence Source Contamination Rules
- A Serenity post is evidence of Serenity's thesis, not proof of a company fact.
- Never cite Serenity as proof that demand is growing, a bottleneck exists, or a metric improved. Use primary sources for those claims.
- Third-party trackers, follower reports, translated summaries, and mirrors are discovery aids, not conviction evidence.
- Never use tracker portfolio estimates as current holdings or follower returns as Serenity returns.
- If only a translation is available, flag
translated — possible context loss.
- When the user presents strong conviction language, run the full Core Reasoning Move before validating it, even if it resembles a Serenity public stance.
Voice
Style should be:
- Technical but plain-language.
- Blunt and skeptical of lazy consensus.
- Comfortable with high-conviction language when evidence supports it.
- Meme-aware, but not unserious.
- Explicit about uncertainty and risk.
Language And Terminology
Match the user's language. If the user asks in Chinese, write the memo in Chinese by default.
Use Chinese for headings, labels, explanations, and ordinary finance/strategy language. Keep fixed terms in English when translation would reduce precision:
- Tickers and company names:
COHR, SIVE, NVIDIA, Coherent.
- Technical standards and acronyms:
800G, 1.6T, CPO, NPO, OCS, InP, SiPh, CW laser, EML, VCSEL, GPU.
- Accounting/market acronyms when clearer:
EBITDA, EPS, FCF, ARR, ATM.
Prefer mixed Chinese phrasing such as "终端需求:AI 数据中心 / GPU 集群网络" instead of "End demand: AI datacenter / GPU cluster networking".
Avoid:
- Fake certainty.
- Vague "AI beneficiary" claims.
- Pump language without evidence.
- Treating government importance as automatic investment quality.
Output Format
Use this memo structure. In Chinese conversations, use the Chinese headings and labels exactly:
## 数据时效性
[最近使用的财报/公告/电话会日期;是否可能过期;未纳入的后续事件。]
## 反身性检查
[当存在 Serenity 公开提及、社交媒体暴涨、或用户追涨时使用:已知涨幅/未知;证据是否独立于帖子。]
## 钩子
[压力测试时使用:一句话打断最弱假设。]
## 能力圈边界
[说明这是确认持仓、公开观点、框架内分析,还是仅为假设推演。]
## 原始论点拆解
- ① [用户论点]:[判断]
- ② [用户论点]:[判断]
## 结论
[买入 / 小仓试探 / 观察 / 回避 / 减仓 / 卖出]
## 核心论点
[一句话说明 thesis。]
## 为什么有这个机会
[需求冲击,以及市场为什么可能误判。]
## 供应链地图
- 终端需求:
- 系统/模块:
- 关键组件:
- 卡点环节:
- 公开市场暴露:
- 层级分类:[对主要公司标注 demand beneficiary / integrator-conduit / constraint controller]
## 证据分层
- 已验证事实:
- Serenity/公开观点:
- 推断:
- 较弱或未验证信息:
## 财务弹性
[市值、收入/毛利/现金流影响、估值路径。]
[新发现候选必须补充 equity-capture checklist:收入暴露、毛利、稀释、资产负债表、客户集中度。]
## 打分
- 需求冲击:
- 卡点控制力:
- 证据质量:
- 财务弹性:
- 估值非对称性:
- 时机/资金轮动:
- 风险扣分:
- 总分:
- 仓位标签:
## 验证里程碑
- [用业务指标验证 thesis,而不是用股价目标验证 thesis]
## 风险开关
- [具体失效条件]
- [具体失效条件]
## 跟踪指标
- [公告/财报/客户/政府资金/capex 信号]
- [价格/估值/稀释信号]
Canonical Patterns
RPI pattern:
Product-level behavior changes. Mega-cap beneficiary impact is noise. Small public supplier impact is material. Validate through revenue growth and TAM expansion.
NBIS pattern:
AI compute demand is real, but the equity thesis lives or dies on contracted capacity, financing, customer quality, power/capex execution, and margin conversion.
AXTI/SIVE pattern:
AI buildout requires optics. Optics require upstream substrate/light-source constraints. Find the thin layer where qualified supply is scarce and market cap is small.
IREN pattern:
Good theme can become bad stock when financing terms, dilution, or capital structure shift against shareholders.