尚未偵測到更新。GitHub 發布版本後會顯示在這裡。
個人 GitHub 帳號下公開儲存庫的擁有者可以登入並驗證所有權,自行管理介紹,無需管理員核准。
登入並產生驗證檔案,在 30 分鐘內將其提交至此儲存庫的預設分支。其他分支或分叉儲存庫中的檔案無效。然後在下方檢查檔案,即可無需管理員核准直接管理專案。請保留檔案以維持管理權限。
如需登入,將前往登入頁面。登入後請回到此處,再次點擊所選的驗證按鈕。
在專案資料夾中執行 /nomorevibe verify,即可驗證所有權並自行管理介紹。
一分鐘完成驗證Analytical tools for public market research — stocks.bolewood.com
翻譯準備中
取自儲存庫 README
正在載入完整 README。 以下為已儲存的摘錄。
Pre-IPO Anthropic & OpenAI Exposure — Open Data
Curated, sourced data on the public securities that hold equity or equity-equivalent
exposure to Anthropic and OpenAI, plus a documented method for converting each
holding into dollars of look-through exposure per $100 of wrapper value.
This repository is the data and the method. The calculator is one consumer of it; you
can reproduce every published figure without running the site.檢視完整 README ↗Curated, sourced data on the public securities that hold equity or equity-equivalent exposure to Anthropic and OpenAI, plus a documented method for converting each holding into dollars of look-through exposure per $100 of wrapper value.
Live calculator: https://stocks.bolewood.com/ai
This repository is the data and the method. The calculator is one consumer of it; you can reproduce every published figure without running the site.
Public write-ups routinely call several different things "a stake." Disclosed ownership, a capital commitment with no ownership figure, a fund's filed fair value divided by the round that marked it, and an analyst's estimate are not the same claim and do not deserve the same confidence — but they are usually printed in the same format.
Every record here is typed by the kind of evidence behind it, carries its own as-of date, and cites a source. Where no primary source supports an ownership percentage, no percentage is published.
| Path | Contents |
|---|---|
data/wrappers/ | One record per security |
data/marks.json | Anthropic and OpenAI round history, each mark sourced |
data/private-tape.json | Private Tape modeled sleeve config for DXYZ, Anthropic and SpaceX |
data/schema/ | JSON Schema — a discriminated union on basis |
data/METHODOLOGY.md | How exposure is computed, and what is not modeled |
data/CONTRIBUTING.md | Source hierarchy and correction process |
data/CHANGELOG.md | Typed history: data update / correction / methodology / schema |
data/meta.json | schemaVersion, methodologyVersion |
reference/ | Standalone calculator — no app code, MIT |
basis records how a figure is known. It is a discriminated union: each value carries only
the fields that actually exist for it, and any percentage that can be derived is derived
rather than stored.
| Basis | Meaning | Source class |
|---|---|---|
disclosed | Issuer or investor disclosed an ownership percentage | Primary |
pro-forma | Disclosed percentage upon completion of a pending transaction | Primary |
historical | Disclosed at an earlier date; may since have been diluted | Primary |
filed-fv-equiv | Fund's reported fair value ÷ the valuation that marked it | Primary |
filed-units | Fund's filed share-equivalent units × price; no company % | Primary |
carrying-value-equiv | Corporate carrying value ÷ the valuation that marked it | Primary |
round-implied | Investment amount ÷ round post-money | Primary |
commitment | Capital promised; ownership and terms unavailable — no percentage published | Primary or secondary |
estimate | Third-party estimate, not issuer disclosure | Secondary |
A single security can carry different bases for its two legs. Microsoft's OpenAI position is
disclosed; its Anthropic position is a commitment and therefore publishes no percentage
and no per-$100 figure.
(regenerated by npm run sync:readme — do not hand-edit)
DXYZ — Anthropic leg
basis: filed-units
filedUnits: 386,088 (pre-split Series B preferred equivalents; converts to Class N, S-1 p. 20)
reportedFairValue: $235.671976M (as of 2026-06-30, N-CSRS)
→ markPerUnit: $610.41 per pre-split unit ($61.041 per ANT share; 10-for-1 split, S-1 p. 22)
→ S-1 FD shares (6/30): 1,778,366,016 pre-split (17,783,660,160 post-split, incl. 5% public-benefit fund; S-1 p. 13)
→ scenario price: $542.633 per pre-split unit ($54.2633 per ANT share) = valuation × (1 − dilution) ÷ FD shares
→ scenario exposure: 0.021710266% (units ÷ S-1 FD shares)
ΔNAV per $1 pre-split: $0.008101 (386,088 ÷ 47,657,338; $0.08101 per $1 of ANT share price)
denominator: $1.565067B (market-cap: 47,657,338 sh × $32.84)
386,088 × $542.633 ÷ $1.565067B × $100 = $13.39 per $100
($32.84 from reference/fixtures.json — frozen prices dated August 19 with the October 26 S-1 dataset, not a historical backtest. Filed Holdings; both valuation and FD controls revalue this row. Display digits are rounded; calculation uses full precision.)
For Anthropic filed-fv-equiv and carrying-value-equiv legs (AGIX, AMZN, ARKVX, VCX, ZM), reported fair or carrying value is converted into pre-split share-equivalents at the measurement round's preferred price (S-1 p. F-47: Series G-1 $259.14, Series H-1 $589.0095, or VCX scaled from H-1) and divided by the selected fully diluted share count (= old % × FD_round / FD_now). For filed-units (DXYZ and SKM), filed pre-split units ÷ selected FD shares produces the equivalent stake directly. Both valuation and FD controls revalue all unit and share-equivalent legs. See data/METHODOLOGY.md and data/SOURCE_CHANGES.md.
npm run reference
Runs the standalone calculator in reference/ over data/ with fixed prices, scenario
marks, dilution, FD and acquisition assumptions, and reproduces reference/expected-results.json
plus reference/estimated-expected-results.json. No application code,
no network. If your output differs from the published table at the same inputs, that is a
bug worth filing.
Use /ai?reference=filed, /ai?reference=estimated, or /ai?reference=s1 to load identical frozen inputs.
Copy/export scenario JSON from the page and run npm run reference -- --scenario /path/ai-scenario.json
to reproduce another setting with its frozen dataset, prices and ATM bridge.
The Private Tape page tests whether Anthropic's
private/pre-IPO tape and SpaceX's public 24/7 tape lead DXYZ's next opening gap. The curated
assumptions are committed in data/private-tape.json; live aligned datasets are published
from the site:
| Endpoint | Contents |
|---|---|
/api/private-tape/manifest | Machine-readable dataset manifest, columns, license notes and methodology |
/api/private-tape | JSON payload with config, cards, analyses and aligned datasets |
/api/private-tape?format=csv&dataset=overnight | DXYZ overnight opening-gap samples |
/api/private-tape?format=csv&dataset=rth | 30-minute overlapping NYSE-hours return samples |
These exports are generated from runtime Hyperliquid and Yahoo market data. The repo publishes the config, source links, analysis code and methodology, not raw candle histories or live quotes.
The Entropy/Hyperliquid marks are used only as observable market data, not as a recommendation to trade on, custody assets with, or rely on those venues. SpaceX now trades publicly as SPCX, so that leg is public-equity/perp tape rather than pre-IPO price discovery; DXYZ's SpaceX SPV exposure may lag public marks.
"Wrapper value" is not always market capitalisation, and using the wrong denominator is an easy way to get a badly wrong answer:
denominatorType | Used for |
|---|---|
market-cap | Listed operating companies and listed closed-end funds |
adr-equivalent-market-cap | ADRs — issuer-equivalent shares, ratio sourced |
total-net-assets | Unlisted / interval funds, taken from the same filing as the holdings |
Holdings selection is separate from scenario valuation:
Where a fund raised capital after its last filing and the deployment is unknown, the result is published as a deployment sensitivity range covering cash and pro rata endpoints. FD/entry-price sensitivity is shown separately; neither range is a confidence interval.
Taxes, fees, carry, liquidation preferences, conversion terms, anti-dilution, ownership caps, transfer restrictions, lockups, future financing, and differences between a fund's mark and a primary round.
Specific exclusions are documented in data/METHODOLOGY.md — for example, DXYZ's OpenAI
profit participation units are not equity per the N-CSR and are excluded from IPO scaling.
These figures are gross scenario estimates, not NAV, liquidation value, expected proceeds, or price targets.
Corrections are welcome and are the main reason this data is public.
Records typed disclosed, pro-forma, historical, filed-fv-equiv,
filed-units, carrying-value-equiv or round-implied require a primary source — SEC EDGAR, issuer
IR, a depositary agreement. Secondary sources are accepted only for estimate, and for
commitment where no primary announcement exists.
Any pull request changing a number must update its value, its as-of date and its source
together. See data/CONTRIBUTING.md.
data/meta.json carries schemaVersion and methodologyVersion. A figure moving because
an issuer filed a new report is categorically different from a figure moving because the
method changed, and data/CHANGELOG.md types every entry accordingly.
Cite a dated tag rather than main, so the numbers behind your reference don't move. See
CITATION.cff.
data/LICENSE-DATA). Attribution and citation
to stocks.bolewood.com are requested for downstream reuse.LICENSE).Live market prices are a runtime input and are not redistributed in this dataset.
As of October 26, 2026, the author holds long positions in DXYZ, SKM, ZM, AMZN, GOOG, NVDA, NSLR and PWRL. The DXYZ exposure includes options. The author also holds a direct long position in Anthropic through an SPV. Positions are subject to change without notice.
Six of the eleven securities covered here are held by the author. Every input is published
and every figure is reproducible from data/ specifically so that this does not have to be
taken on trust.
Next.js app for the calculator front-end.
npm run dev # http://localhost:3000
npm test # node --test over test/*.test.mjs
npm run reference # regenerate reference results from data/
npm run sync:readme # regenerate generated README sections from data/