Scope · AFS
Of Stanley Druckenmiller / Duquesne, for you. Cites the locked monthly. Does not replace §00. Not a buy list.
Evidence base
Scout corpus druckenmiller-operating-system-2026-Q2. Primary: three EDGAR 13F-HR information tables. Aggregators used only as cross-check.
August snapshot
Your 11, the retro, and the September plan: /2026-08. Pair case study: Aschenbrenner. Cycle engine: BTC bear. Coefficient stays 0.
Verdict — reverse the vehicle, keep the job
What happened: Q1→Q2 he re-risked the visible equity book (~$3.38B → $5.21B), fully exited 23 names, opened 48, and bought Alphabet back after a quarter of zero. What went right: regime first, vehicle second; full exit, not a lingering trim. What this filing cannot score: cash, shorts, gold, FX, rates — it is a delayed photograph, not the whole book.
Not a holdings list. Steal the operating system. Pair: Aschenbrenner ran adjacent AI-capex exposure with opposite structure — and, at Q1 2026, opposite semis positioning (Duquesne long TSM/STM; SA held puts on TSM and other semis). Do not flatten. Your 11 live on the August monthly.
What happened
Three consecutive 13F-HR filings. The story is sleeve rotation, not 95 line items.
| Window | Job | Print [A] |
|---|---|---|
| Q4 2025 | Held Alphabet | 385,000 sh · $120.5M |
| Q1 2026 as-of 31 Mar · filed 15 May | Full GOOGL exit · book ~$3.38B / 70 rows | Zero Alphabet |
| Q2 2026 as-of 30 Jun · filed 14 Aug | Re-risk · GOOGL rebuy · healthcare kept as anchor · US cyclicals and listed-crypto satellites new | $5.21B / 95 rows · GOOGL 336,300 sh · $120.2M |
Illustrating moves (not a buy list)
- Alphabet: full quarter off, then same-dollar re-entry. Share counts are [A]. “After earnings confirmed the AI-capex job” is Scout [Inference] on [B] framing, not a filing fact.
- Alcoa cut ~90% while Cleveland-Cliffs and Fluor were added — reshoring job kept, vehicle swapped.
- Healthcare dollars held (Natera still #1 at 16.6%); weight diluted because the rest of the book grew faster.
- Broadcom, Intel, Micron, Lattice fully exited — AI stack kept via TSM / STM / new AMD, not a GPU consensus tape.
- Listed-crypto wrappers (Bitdeer, Hut 8, Riot, IREN, Hyperliquid Strategies) appeared as a ~2.9% satellite, all new in Q2. 13F cannot hold spot BTC.
- Call-option notional ~16% of reported value (RSP, IWM, EWZ, AMZN…). 13F prints underlying notional, not premium — capital-at-risk is much smaller.
- 23 full exits and 48 new in one quarter — complete reverse, not trim-to-zero-conviction hold.
July 2026 AI tape is after the 30 Jun as-of. This photograph does not say whether AMZN/TSM adds survived July.
Why
Nine process rules from the filings. Size is attention; stale thesis gets a full exit; the regime outlives the ticker.
| # | Rule |
|---|---|
| 1 | Concentrate — top 10 ≈ 46% of the visible book. Natera alone 16.6%. |
| 2 | Concentration is the risk control (attention), not a forgotten 40-name tail. |
| 3 | Reverse completely when the slot is stale — 33% full-exit rate Q1→Q2. |
| 4 | Regime first, vehicle second. GOOGL round-trip is the clean print. |
| 5 | Listed calls press a basket/regime; single-name conviction is cash equity. |
| 6 | Healthcare is the sleeve he does not rotate out of. |
| 7 | Second-derivative US housing / airlines / construction when AI-capex conviction is already funded. |
| 8 | Sub-1% names are scouts, not a directional tape. |
| 9 | Crypto via listed equity proxies, satellite, late in this risk-on quarter — not core, not spot. |
Right / wrong
Right on this photograph
Unforced reverse. Same job, new vehicle. Family-office book with no public prime-broker spiral in this window. Healthcare left alone while he re-risked elsewhere.
Not scored / can be wrong later
13F is 45 days late, long-only US-listed. Options notional overstates risk. Intra-quarter timing is a gap. Q2 re-risk into AI-infra before July is a timing bet this filing cannot grade.
Do not flatten against Leopold as “he was smarter.” Contrast is structure and reverse-without-a-gun, not a skill score.
Against our data
No live Layer 1 / blotter contest is poured here. A named-operator clause belongs on the monthly only when a P-* or WL is actually matched or contested — our row first.
No live August line contests this photograph. Scout holds unused stubs in the corpus (AI-capex, reshoring, listed-crypto adoption) — none maps to a current P-* or WL. Sentence order stays (1) our tape → (2) the job → (3) delayed Duquesne, tagged secondary.
Dataset gaps
| Gap | Ours? |
|---|---|
| G1 | Cash, shorts, gold, FX, rates — 13F-blind. Never-ours as a Duquesne blotter. Layer 1 already has gold/rates our series — that is our tape, not his book. |
| G2 | Option premium vs share-notional. Candidate: if we ever store options, pin premium not 13F notional. |
| G3 | Intra-quarter fills (when GOOGL actually left). Never-ours for any 13F filer. |
| G4 | Official sector tags — Scout sleeves are inference. Do not ingest as GICS. |
| G5 | Post-30 Jun July path. Next photograph: Q3 13F due on or before 14 Nov 2026. |
Not a ticker shopping list. Do not add Duquesne names to Cloud QUANT MCP.
Appendix — reconstructed sleeves
Scout classification, not audited GICS. Lag banner: as-of 30 Jun 2026, filed 14 Aug.
| Sleeve | Q2 % | Q1 (same bucket) |
|---|---|---|
| Healthcare core | 31.4% | ~36.3% |
| US broadening | 28.2% | lower — many names new |
| AI / semis stack | 21.0% | present, smaller dollars |
| EM | 14.0% | similar weight, different names |
| Listed-crypto wrappers | 2.9% | 0% — entire sleeve new |
Top-20 tickers stay in the Scout corpus, not on this page.
Notes
- Case study — of Duquesne / Druckenmiller, for you. Not a Profile.
- Special report — the BTC cycle brief only.
- Sources — Scout corpus
druckenmiller-operating-system-2026-Q2· EDGAR accessions 0001536411-26-000002 / 000004 / 000006.