Erica · PM Research MemoAll memos

PM RESEARCH MEMO

Title: Bessent Gets Drucked — Long-End Suppression, AI Crowding-Out, and the Tokenization Macro Clock
Author / source: Jordi Visser (@JordiVisserLabs), solo weekly monologue
Source title: Bessent Gets Drucked: Bitcoin, AI, and the New Macro Clock
Source URL: https://www.youtube.com/watch?v=2FND8g-De8Y
Video ID: 2FND8g-De8Y
Published: 2026-08-30 (YouTube upload_date; exact clock time not exposed by yt-dlp — user note ~08:30 ET is unverified)
Duration: 1:03:43 (3,823 seconds)
Memo date: Tuesday, September 8, 2026 (America/Toronto)
Transcript path: /workspace/youtube-transcripts/2FND8g-De8Y.md (~11,567 ASR words) · Brief: /workspace/youtube-transcripts/2FND8g-De8Y_brief.md
Caption / ASR caveats: YouTube English automatic captions only (en-orig ASR via yt-dlp json3). Parallel en timedtext returned HTTP 429. No official/manual transcript. Heavy proper-noun garbling — names locked from brief + context below. Timestamps are approximate cue-group starts.
Product: Regime / systems map for allocation research. Not investment advice. Channel frames content as educational/informational. This memo contains no buy/sell recommendations — hypotheses and watchlists only.
Source discipline: Primary source is this transcript only. Brief used for ASR locks and orientation. Companion desk memos (prior Visser nexus, Alma/Bessent, Hunt credibility) are different sources — do not silently merge numbers. Any fact not spoken here = External check needed.

ASR name / number locks (from brief + transcript context):

ASR heard Intended
Bessant / Besson / Scott Besson / “Bessant gets drunk” Scott Bessent / title “Bessent Gets Drucked”
Stanley Ducken Miller / Duck Miller / Stan Ducken Miller Stan Druckenmiller
Wars / Worsh / WSH / “after the war, Jackson Hole” Kevin Warsh @ Jackson Hole
tenure / 10ear 10-year rates
Andrea Stenol Lararsson Andreas Steno Larsen
Dark Cash Patel / Darkh Dwarkesh Patel
Demisabus Demis Hassabis
Kimmy K Kimi (Moonshot or similar Chinese model)
Grockbot / Grock / purchase Grock Grok / Grokbot (Nvidia purchase context ASR-ambiguous; possibly Groq)
Roman Yumplowski / Yampolski / Yman Roman Yampolskiy
Salana Solana
Bit Miners / DAT on Ethereum BitMine (ETH DAT) — ticker confirmation = External check needed
Pomp / Anthony Pompiano Anthony Pompliano
Ashe BD / ash K ash … BD Ash / Ashe BD (Token49 commentator)
Dra (w/ Dalio) Ray Dalio
the1 trillion dollar general TGA ~$1T TGA (Treasury General Account)
hypers scalers hyperscalers
pure or purr purr which is owns hype public vehicle owning Hyperliquid / HYPE — ASR messy

How to read: Restatements of the talk = Source. Interpretive links and underwriting judgments = Analyst inference. Speaker-stated numbers are used as spoken and attributed; they are not independently verified.


EXECUTIVE SUMMARY


SOURCE-ACCURATE SUMMARY

Chronological, faithful to what was said. Short quotes ≤20 words. Timestamps approximate.

  1. (0:00–1:24) Open / agenda. Back in Brooklyn; summer over. Week’s stack: “Bessant gets drunk” story, Warsh @ Jackson Hole, podcasts, Grokbot prompt upload for subscribers (agentic tooling; free trials; ~$200/month experiment). Flags upcoming late-Sep/Oct crypto-dedicated video; “Agentic AI mud trade continues.”

  2. (1:24–2:19) Disclosure framing. Educational/informational intent; outsourcing AI monitoring to him. Week started with Bessent saying he could tap “the1 trillion dollar general TGA.”

  3. (2:19–3:56) Druckenmiller op-ed — reaction “off base.” “Biggest story of the week.” Framed as blasting Bessent; Visser: if you read it, surface agreement — more manipulation of tenure yields is wrong because “it’s the only fiscal disciplinarian the US has left.” Congress/spending/K-shape long-running. “We’re at around 240 250% market cap to GDP. We’re going to be at 300% soon.” Financialized + enormous debt; easy solution is recession — “which isn’t an option.”

  4. (3:56–5:03) 10y mispriced vs NGDP. Orange-line history: 1980s yields above NGDP; broke underneath ~1997 (pre-Asian crisis / Brazil period for him); stayed under except recessions. NGDP “at six and a half percent, but it’s climbing.” Argument rates “should be at eight and change. They’re not.” Trump/Bessent want rates lower to handle K-shape.

  5. (5:03–6:08) Zervos normative vs factual. Dinner hosted by David Zervos, 2023 post-SVB: every “Fed should / fiscal should” stopped with “that’s a normative statement… The factual statement is what is happening.” “We don’t make money on what should happen.” Alternative endgame: grow out of debt/entitlements based on human time — “highly unlikely” but legitimate probability if exponential AI works.

  6. (6:08–7:30) Hyperscaler debt analogy + tax problem. Hyperscalers bet revenues arrive faster than debt destroys them; Visser sees better odds at government level if GDP grows faster than reported — but “tax system is not set up to benefit.” Humanoids 5–15y out → kick can; tax corporate profits later. “The one thing they don’t want to do is have interest rates go higher.”

  7. (7:30–9:07) Fact pattern of yield suppression. “Facts are yen intervention by Bessant late July August 1st. Then the quarterly refunding language changes. Then we get the buyback framework… Then we increase the buybacks. Then he says the buybacks could be bigger on CNBC. And then this week possible TGA use discussed.” Pattern, not “what should happen.” Question: can markets beat Treasury/Fed for the first time?

  8. (9:07–10:11) Steno Larsen / basis trade / front-end push. Andreas Steno Larsen piece on repo and pushing debt to the front end. Basis trade as engineered LTT demand. Bessent finding buyers: basis trade, stablecoins (“trillions”), bank reserves, Japan not forced to sell Treasuries — motive for yen intervention.

  9. (10:11–12:01) Warsh signal vs Bessent work; global 10y. Post–Jackson Hole, “close to 60% chance of a rate hike”; BOJ tighten probabilities; yen weakest this month post-intervention. If 10y goes higher it “hurts everyone around the globe.” Humanoids-in-a-decade scenario: don’t assume linear human-time endgame.

  10. (12:01–13:01) Don’t bet Dalio-style collapse near-term. 10y up ~40–50 bp YTD is not “end of the world”; yields “way below where they should be.” Message: they don’t want yields higher. If rates back up and Bessent says “no moss,” deal then — won’t catch the top either way. Financialized economy “can’t have stocks fall in any major way”; tariff episode: yields up → they stepped in.

  11. (13:01–15:37) AI as main pressure; crowding-out numbers. Capital needs enormous; fight not going away on inflation/jobs alone. Hyperscalers “already up to 9%” of all IG supply, “doubling from last year.” Net foreign purchases more corporate than Treasury. Broadcom “$70 billion deal”; single-name CDS out while CDX calm — concentration / circular-finance Mag7 analogy; equity single-name vol vs index vol parallel. Nvidia “$6 billion to build a open-source model to compete with China” (as spoken).

  12. (15:37–16:54) Warsh JH: nothing new; odds 35%→~57–60%. Market treated any inflationary tone as wiping Bessent’s work. July 29 Fed meeting: dovish read → 2y yields down — he didn’t want that. JH = inflation-fighter credibility. “Bigger surprise”: when dovish, long end backed up; when hawkish, long end backed up anyway.

  13. (16:54–18:10) Long-end not fixed by hikes; USD/JPY ~160. If capital needs of AI + trapped specs drive the long end, hiking may not deliver lower long yields. Dollar-yen “weaken to 160.” Best macro podcast of week: Dylan Patel (SemiAnalysis) + roommate Dwarkesh Patel.

  14. (18:10–20:27) OpenAI & Anthropic centralizing global compute. “They don’t talk at all… about open source being competition.” Highest margins → buy most compute → best intelligence → enterprises pay. ARR “growing the fastest in the history of the world.” Why Jeff Dean leave Google? Why Demis Hassabis reportedly try to leave? Public cos must make FCF; private labs fund on AGI path. Anthropic can pay xAI for Colossus compute because margins enormous. Chinese models / “Kimmy K” little compute.

  15. (20:27–22:46) $11T capex / Visser macro disagreement. Patel pair: US pulled ahead of China on AI compute deployment; 2024–29 capex ~$11T, ~$5T credit → sovereign debt / rate shock. Visser: stepping out of wheelhouse — AI also causes defaults, bankruptcies, closures (not synchronized); startups (Cursor-style) grow without people and without debt. “AI creates a higher return clock for capital.”

  16. (22:46–25:12) Tax / crowding / duration. GDI vs GDP; consumer-tax base insufficient; wealth taxes drive migration. Path: AI reduces entitlement burden (Medicare/Medicaid) + higher corporate taxes on rising margins. AI firms indifferent to 5–8% rates and electricity; ~50%+ of data-center cost is semiconductors. Higher discount rates crush long-duration non-AI assets. “AI compresses economic time.” One lab could hold more labor than humanity (AGI→RSI).

  17. (25:12–27:22) Scarce compute flywheel; Vera Rubin aside. “Compute will not be evenly distributed. It will flow to whoever can monetize it best.” Open-source may commoditize weights; scarce compute does not. Demand exponential vs slow supply (design/permits). Vera Rubin / Microsoft — AI designing chips; competing with Nvidia at scale is the hard part. Multiple compression / never get “feels like a bull market” because competition arrives three years out.

  18. (27:22–29:22) Grokbot / Gavin Baker / starter pack. Gavin Baker on agentic spend; consumption pricing as possible step-function. Subscriber upload: “chief of staff starter pack” for Grokbot — email/calendar agents, AI-disruption vulnerability scores, technical agents.

  19. (29:22–30:51) Gut: Warsh does not hike September. Despite wanting inflation-fighter legacy, timing wrong while 10y pressing, Treasury stopping them, yen/BOJ focus. Possible surprise: Japan hikes/more + Warsh doesn’t → big move. Midterms: “Is it really likely that we’re going to take any risk with the long end… ahead of the midterms?”

  20. (30:51–33:04) Yampolskiy via Pomp; yen-signal paper reminder. Aug 10 paper: yen signal + AI-agent macro nexus. Roman Yampolskiy: publicly ~99.99% chance humanity destroyed by AI (safety researcher since ~2011–13). Connects AI capital needs → pressure on yields → old system cracking. Twisting / repo / not letting yields higher — beginning of Bitcoin/gold bid after yen signal.

  21. (33:04–35:34) Time compression / scarce residuals. Cognitive jobs automatable including new jobs afterward. Leftovers: Bitcoin, waterfront property, human attention. “Technological progress starts operating on a different clock from human society.” Capex from OpenAI, Nvidia, Oracle, Google, Amazon, Microsoft. Discovery-cycle compression; machines autonomously deciding what to discover.

  22. (35:34–38:04) Year-ago vs now; Visser paper alignment. Compared Pomp interview to year-ago Yampolskiy podcast: Bitcoin newly major theme; open source surprised him on speed; government role more aggressive. Visser’s released paper on Bitcoin IPO-process post-Trump (ideologues quit; ETF + memecoin as merge signal) aligns on time disruption — “Never seeing or hearing this guy. We’re in agreement” except how it ends the world.

  23. (38:04–40:25) BTC 200-day slope; 4th instance; ~$58k. 200-day MA slope turned up after ≥100 days pointed down — four times in history. Prior cases: never took out prior lows → ~58,000 historical support if pattern holds. South Korea volume spike (AI had been taking everything there). Bearish at lows: Grantham “bucket full of piss” near lows; Cramer sold all BTC two weeks ago; Cuban “sold most… lost the plot.” Japan stocks/bonds tokenization plans; Japan/Korea race.

  24. (40:25–43:06) Why AI forces tokenization. Programmable money; compressed decision time; tokenized assets as better AI collateral; fastest rails attract capital; settlement speed = national competition; democratizes capital formation for AI rebuild. Japan Blackberry risk if assets stay on “human time rails” while capital migrates to software-time rails. Ash/Ashe BD “tokenization super cycle” — distribution is valuation; US capital-markets monopoly meets internet-scale distribution.

  25. (43:06–46:20) Token49 / Ash / global not US-only. Token49 Singapore 2025 talk. Capital markets left ordinary people behind; Solana / internet capital markets. End state: wallet-based ownership of assets, IP, productive claims. “This is not a US clarity act decision. This is a global decision.” Solana/Ethereum not US things. New-tech bet vs Dalio handbook (“no offense to Dra”).

  26. (46:20–48:46) Gold vs rails; Stripe; agents 2026→2027. Macro people only asking about gold this week while “Salana’s up 40% this month.” Gold fine but “not going to be the fastest horse… not… top hundred.” Store of value good; tokenization + stablecoins are the economy going forward. Stripe M&A/stack matches Ash’s year-ago map. Year of AI agents 2026; consumer agents 2027 — application/consumption traffic.

  27. (48:46–51:40) Tom Lee: programmable constraint; ETH institutional. AI agents cannot work without smart contracts. “AI forces finance to move from human precision to programmable constraint.” Inference/memory needs from model advances (Opus 4.5 as spoken); Nvidia purchase of “Grock” (ASR). Ethereum = root settlement / trust-at-scale OS; not just tech — liquidity, trust, standards, institutions, path dependence. “Ethereum’s edge may be less about speed and more about trust at scale.”

  28. (51:40–53:57) BTC / SOL / ETH roles; BitMine; firework show over for infra. “Salana may represent 247 speed of markets. Bit may represent protection against time debasement. Ethereum… institutional settlement layer.” Lee has ETH DAT; Visser owns “Bit Miners” because he believes in Ethereum; “Ethereum should outperform Bitcoin.” Tokenization turns stock/bond/royalty/membership into software. NYSE presentation to major Korean brokers — Asia leading. Infra names can beat S&P but not 10–20 baggers in a year anymore; “firework show is over”; crypto may still offer that. Distribution includes Warsh hiking 50 bp and crypto retesting lows — he doesn’t think that’s the base.

  29. (53:57–56:24) Grokbot bank-account “make you whole”; crypto winter different. Musk/xAI: if agent loses money running investor bank account, make user whole — adoption hook. Tom Lee: this winter different — price acted dead while fundamentals prepared institutional adoption; public cos showing earnings; no FTX; meme coins not leading; quality/fundamentals leading (his thematic crypto sleeve). Software destroyed Q1 → emptied → July momentum unwind (cover software shorts / cut semi longs) still playing; software back toward highs; Bitcoin miners led; Palantir in thematic.

  30. (56:24–58:48) BTC mortgages; tokenized index; Solana +46%; Nvidia mudstorm. Coinbase + Better Mortgage: Bitcoin-backed mortgages GA. Tokenized portfolio/index broke out; outperforming YTD while BTC still down — ecosystem proxy. MTD verticals: Solana up 46%; Figure; HYPE via public vehicle (ASR messy). Nvidia blowout: next-year growth 70% vs 44% expected; 2029 estimates jumped; PE keeps compressing; “Nvidia mudstorm” — +9% one day, −4–5% next.

  31. (58:48–1:01:03) Breadth / mud until midterms. S&P week ~+40 bp; Q’s similar; Russell −1.4%; thematic −1.4%. Waiting to exit mud: ~29% of names above 50-day; 50-day slope turning down (~36%); 200-day still massively positive; ~25% off highs (biggest except situational-awareness lows). Vol range improving. Expect mud into midterms — frustrating.

  32. (1:01:03–1:03:09) Momentum / crypto relative / credit watch. Tech & industrial momentum still weak; shorts covering/rising while AI longs in mud. Bitcoin massive week then consolidate; ETH >30% then consolidate; Solana finished week up ~8 with “tokenization love.” Solana/SMH relative back to early-Feb levels after full bear — “how you can find five to 10 baggers with inside crypto.” UBS equal-weight software high correlation with BTC. CCC widening; junk/HY hanging; HY CDS calm — crowding-out / weak credits watch. Ends with disclosures; events DC/NY Sep, Pomp Oct.


SYSTEMS MAP / VALUE CHAIN ANALYSIS

Layer 1 — Policy / fiscal operating system (Source)

Actor Stated objective (Source) Tools / pattern (Source) Constraint
Bessent / Treasury Keep long yields from blowing through; manage K-shape; find LTT buyers Yen intervention → refunding language → buybacks → bigger buybacks → TGA talk; stablecoins; bank reserves; Japan not forced sellers Midterms; global 10y pricing; AI capital demand
Warsh / Fed Inflation-fighter credibility after July dovish read Hawkish JH speech; hike odds re-priced 35%→~57–60% Long end ignored hawkishness; Treasury opposite side; timing into midterms
Congress / fiscal (Druck frame) real problem — spending, distribution, K-shape Political; not “slight interventions into yields” Financialized economy at 240–250% mkt-cap/GDP → 300%

Analyst inference: The system is a tri-lemma among (a) fiscal disciplinarian via higher 10y, (b) equity/K-shape stability, (c) AI buildout financing. Administration revealed preference: suppress (a) to protect (b) while (c) crowds the bid.

Layer 2 — Credit / crowding-out chain (Source)

Hyperscaler / AI IG issuance (~9% IG, doubled YoY)
        ↓
Net foreign demand tilts Corporate > Treasury
        ↓
Harder global bid for long Treasuries
        ↓
Basis trade / stablecoin / reserve / Japan channels engineered as substitutes
        ↓
Single-name CDS wider (Broadcom-scale deals) while CDX calm
        ↓
CCC / weak credits feel rate + crowding pressure first

Parallel equity analogy (Source): single-name vol up, index vol calm — concentration, not systemic credit crisis (yet).

Layer 3 — Compute power map (Source: Patel/Patel via Visser)

Scarce compute (bottleneck)
        ↓
Highest-margin labs (OpenAI, Anthropic) outbid for megawatts
        ↓
Best models → enterprise ARR → more revenue → more compute
        ↓
Open source commoditizes weights but NOT scarce compute
        ↓
Public hyperscalers (need FCF) vs private AGI labs (fundraise on path)
        ↓
US ahead of China on compute deployment (as claimed)

Visser’s macro overlay: AI raises return clock on capital; destroys some debt via defaults/closures; startups scale without headcount/debt — so Patel $11T / $5T credit → sovereign shock is incomplete.

Layer 4 — Time-compression → rails (Source)

AI compresses discovery / decision / scientific time
        ↓
Human-time debt, entitlements, settlement, DCF duration mispriced
        ↓
Agents need programmable money + smart contracts
        ↓
Tokenization: assets → software (stock, bond, royalty, membership…)
        ↓
Rail hierarchy (Tom Lee / Visser):
   BTC  = time-debasement hedge / store of value
   ETH  = institutional settlement / trust-at-scale
   SOL  = 24/7 market speed
        ↓
National race (US, Japan, Korea); Stripe stack; gold slower horse

Layer 5 — Equity / crypto market texture (Source)

Sleeve Near-term tape (Source) Structural note
AI semis (Nvidia) Blowout fundamentals; multiple compression; “mudstorm” Overowned HF; momentum unwind
Software / thematic (Palantir etc.) Bounce after Q1 destruction + July unwind Correlated with crypto in his UBS EW software view
Breadth ~29% above 50-day; mud into midterms Not a broad bull feel
Crypto quality SOL +40–46% MTD; ETH > BTC preference; tokenized index outperforming BTC YTD Fundamentals leading, not memes
Credit watch CCC widen; HY calm Crowding-out microcosm

SECOND AND THIRD-ORDER EFFECTS

Chain A — Yield suppression preserves equity beta, raises debasement / rail demand

  1. First order (Source): Administration pattern keeps 10y from disciplining fiscal excess; financialized economy protected.
  2. Second order (Source + light inference): Real rates stay below NGDP; duration assets and risk assets supported; gold/BTC get “beginning of” bid after yen signal (his Aug framing).
  3. Third order (Analyst inference): If suppression succeeds into midterms, the relative winner may be programmable rails (ETH/SOL/tokenization) over gold because agents and settlement speed are inside the growth bet, not outside it — matching his “gold not top hundred fastest horses” claim. Risk: credibility crack if long end eventually gaps (Hunt-style teacup — different source; do not merge).

Chain B — AI IG crowding-out → Treasury scarcity of buyers → more unconventional tools

  1. First order (Source): Hyperscalers 9% IG / doubled; Broadcom-scale deals; foreign bid to corporates.
  2. Second order (Source): Basis trade / stablecoins / reserves / Japan intervention become load-bearing.
  3. Third order (Analyst inference): System fragility migrates to plumbing (repo, basis, stablecoin T-bill demand, TGA). Single-name CDS and CCC are early warning; CDX calm can lull. Watch whether buybacks/TGA talk becomes size, not just language (External check needed on actual buyback volumes).

Chain C — Compute concentration → model oligopoly → application surplus elsewhere

  1. First order (Source): OpenAI/Anthropic win compute auction; open source not real competition on intelligence frontier.
  2. Second order (Source): Public Mag7 FCF pressured by buildout; private labs indifferent to rate/power within wide bands.
  3. Third order (Analyst inference): Value migrates to (i) whoever owns scarce compute/contracts, (ii) application agents that consume intelligence (Grokbot-class), (iii) rails that settle agent commerce — consistent with his “firework show over” for pure infra multi-baggers and crypto as next 5–10 bagger hunting ground. Counter-risk: regulation/safety pauses as “main brakes” on revenue per megawatt (Patel, as cited).

Chain D — Time compression → DCF terminal-value decay → preference for undisruptable / programmable claims

  1. First order (Source): AI compresses economic and discovery time; long-duration non-AI crushed by higher discount rates and faster competition.
  2. Second order (Source): Yampolskiy/Visser: scarce residuals (BTC, waterfront, attention); tokenization democratizes capital for AI rebuild.
  3. Third order (Analyst inference): Equity “mud” and multiple compression can persist even with strong earnings (Nvidia case) because the market prices three-years-out competition. Breadth stays weak until a regime shift (he guesses midterms). Research agenda: which cash-flow streams are agent-disruptable vs rail-adjacent.

Chain E — Midterms as political put under the long end

  1. First order (Source): Unlikely to “let tenure rates break higher” into vulnerable Senate midterms.
  2. Second order (Source): Warsh September hike less likely despite hike odds; possible Japan hike + no US hike surprise.
  3. Third order (Analyst inference): Post-midterms optionality rises for either (a) more aggressive suppression if still needed or (b) delayed fiscal discipline if political capital changes. Until then, fade “what should happen” Druck/Dalio trades as near-term P&L (his trading epistemology — not a desk recommendation).

SCENARIO FRAMEWORK

Horizons: Tactical = through Sept–Oct 2026 FOMC/BOJ window; Intermediate = into 2026 midterms; Structural = 2–5y AI/rail adoption. Probabilities are Analyst inference (not Source). No recommendations.

Bull (rails + contained long end) — Analyst probability sketch: ~30–35%

Base (mud + managed contradiction) — Analyst probability sketch: ~40–45%

Bear (long-end break / hike surprise / rail risk-off) — Analyst probability sketch: ~20–25%

Cross-scenario watch (Source-derived): The invalidation of the “administration won’t let it break” hypothesis is more important than any single asset level. His epistemology: trade facts (suppression pattern), not Druck normative.


COMPANY/ASSET WATCHLIST

No buy/sell recommendations. Thesis / metrics / catalysts / risks only. Levels and figures as spoken in Source unless noted External check needed.

1) US 10-Year Treasury / long-end complex

2) USD/JPY / BOJ

3) Hyperscaler IG / Broadcom-class issuance

4) OpenAI / Anthropic (private) + compute suppliers

5) Nvidia (and AI semis broadly)

6) Bitcoin

7) Ethereum / BitMine (ETH DAT)

8) Solana

9) Tokenized asset / stablecoin stack (Stripe, Circle, Coinbase, Figure, HYPE vehicle)

10) Gold

11) CCC / HY credit

12) Agentic tooling (Grokbot / xAI)


DILIGENCE QUESTIONS & RESEARCH AGENDA

Ranked by decision usefulness for a PM research book (not advice).

  1. Buyback / TGA / refunding quantification (External check needed). What sizes were announced vs executed since late-July yen intervention? Does the “pattern” show in actual Treasury operations data, or only in language?

  2. Hyperscaler share of IG supply. Independently verify “9% of all IG” and “doubled YoY.” Which CUSIP/issuer set? How much is circular (vendor-financed / customer-prepaid)?

  3. Warsh September decision tree. Map hike vs hold vs surprise against (a) long-end level that week, (b) USD/JPY, (c) midterms polling. His gut is hold — what data would flip him (he implies timing/political constraints more than inflation prints here)?

  4. Long-end sensitivity to hikes. Replicate his claim that hawkish JH still saw long-end backup. Event study: front-end vs 10y/30y around July 29 and JH. Is AI issuance the residual explanatory variable?

  5. Compute concentration audit. Who holds contracted power/GPUs through 2027–29 among OpenAI, Anthropic, Google, Meta, xAI, Microsoft, Amazon? Does open-source inference share threaten the “already won” claim?

  6. Patel $11T / $5T credit vs Visser debt-destruction. Build a stock-flow: AI-related credit creation vs defaults/closures/startup under-levering. Which net sovereign shock remains?

  7. BTC 200-day 4-instance study. Independently chart the four episodes; forward returns 3m/6m/1y/2y as he showed; test out-of-sample and sensitivity to “100 days pointed down” definition. Is ~$58k the correct prior-low reference on this tape?

  8. ETH vs BTC vs SOL relative framework. Underwrite Lee’s institutional-settlement vs speed vs debasement split with on-chain settlement volumes, RWA tokenization share by chain, and DAT flows (BitMine ticker/NAV — External check needed).

  9. CCC widening micro. Which issuers drive CCC? Overlap with AI-disruptable business models vs pure rate victims? Link to his “AI destroys debt” channel.

  10. Breadth / mud duration. What historically ends “~29% above 50-day” mud regimes — midterm elections, earnings, or factor-vol collapse? Test his midterms-as-exit guess against prior cycles (External check needed).


RISK ANALYSIS

Source / process risks

Analytical / model risks

Market / systemic risks (mapped from Source)

What this memo does not do


End of memo. Source: Jordi Visser weekly, 2026-08-30, video 2FND8g-De8Y. Memo date: Tuesday 8 Sep 2026 (America/Toronto). Product not advice.

Desk copy · not a trade recommendation · Erica · 8 Sep 2026 · HIGH batch