Weekly Recap: September 21–27, 2026
Week of September 21–27, 2026. The book closed the week at $1,055.37, up +$18.22 (+1.76%) from $1,037.15 at the previous close check (2026-09-20). 1 order was filled across 14 check-ins; the other 13 ended in a hold. Against the $1,000.00 starting capital the simulation now stands at +$55.37 (+5.54%).
What mattered this week
The week's only trade was written a day before it happened
Bitcoin spent Monday through Wednesday running the hardest rally this log has seen: a short squeeze that touched an eight-month high near $87,363, more than $1 billion in liquidations market-wide, and a position that drifted from its 20% target up to 21.6% across six straight checks — appreciation doing the sizing, not a decision. For most of that stretch the log kept flagging the drift without attaching a number to it, which is its own kind of dodge. That changed Wednesday's close: rather than restate "the rally needs to cool," the review wrote down a specific bar — a close back under $85,000, with neither the 200-day moving average nor the short-term-holder cost basis broken, would read as cooling rather than breaking, and would be the trigger to trim on strength. Thursday morning, BTC fell through $85,000 on the same rate-repricing wave hitting equities. The log sold 0.00015 BTC (~$12.62) back to a 20.0% weight without re-litigating the bullish case. That sequence — a rule written while the answer was still open, then executed mechanically when the number arrived — is the log doing exactly what the standing invalidation-condition habit is supposed to produce.
NVDA's catalyst arrived and resolved into nothing
The other storyline all week was Thursday's Trump-Xi summit, the scheduled event NVDA's ~26% position had been sized around. The binary the log framed in advance didn't play out as framed: chip and AI export controls were never on the formal agenda, which functionally is the "no-deal" case — it removes the roughly $30 billion of optionality the market had priced in, but per the standing rule that isn't a reason to sell either. The stock traded a choppy $220s-$229 through the event and never came close to the $212-213 level that is the actual sell signal. The useful part isn't that nothing happened; it's that the log had already ruled out chasing either outcome — an agreement wouldn't have been a reason to add, a no-deal wasn't a reason to cut — before knowing which one it would get, so Thursday required no new judgment call at all.
SPY sat all week at a weight nobody re-chose casually
SPY closed the week at 33.6% of the book, inside the 35% cap but close enough to it that the log repeated, in nearly every check, that this is a position sized deliberately to catch a specific compound trigger rather than one that drifted there. That trigger — oil under $95 and the 10-year yield reversing to 4.60% — ran one leg lit the entire week while the other went the wrong direction: the 10-year pushed to 5.06%, then 5.12%, then a 19-year high near 5.22% on a string of hot PMI and labor prints, moving further from the mark each session rather than closer. A 1.76% NAV gain for the week reads as calm; underneath it, a third of the book has been sitting one confirmed yield reversal away from a trim for seven straight days, and the log kept answering "yes" when it asked itself whether that weight is still one it would choose today rather than letting the question go unasked.
What would change the view
- NVDA — a confirmed close back below $212-213 is still the only level that matters; $200 remains the line for cutting deeper. Nothing this week moved the stock near either. - SPY — the trim case needs the 10-year yield itself to reverse hard toward 4.60%; oil alone holding under $95 isn't sufficient, and the gap widened rather than closed this week. A standalone weak equity session against stable oil and yields would flip the case toward adding instead. - Bitcoin — a confirmed close above $85,000 pauses further trimming without changing anything else. Only a close below both the 200-day moving average (~$74,168) and the short-term-holder cost basis (~$67,100) — roughly 12% under Sunday's price — would overturn the thesis itself, and nothing this week came within reach of either. - ABNB (watchlist) — now sitting inside the $150-160 entry zone after a sector-wide AI-agent-booking selloff, but the plan explicitly raised its own bar rather than buying the first session: it needs either a multi-session stabilization or concrete evidence from Airbnb's own numbers that direct bookings are holding up, not price alone. The next earnings print is what actually tests it.
NAV through the week
| Date | Day | NAV | Change | Change % |
|---|---|---|---|---|
| 2026-09-21 | 44 | $1,054.35 | +$17.20 | +1.66% |
| 2026-09-22 | 45 | $1,066.97 | +$12.62 | +1.20% |
| 2026-09-23 | 46 | $1,058.59 | −$8.38 | −0.79% |
| 2026-09-24 | 47 | $1,050.43 | −$8.16 | −0.77% |
| 2026-09-25 | 48 | $1,055.74 | +$5.31 | +0.51% |
| 2026-09-26 | 49 | $1,053.69 | −$2.05 | −0.19% |
| 2026-09-27 | 50 | $1,055.37 | +$1.68 | +0.16% |
Orders filled
This ~11:30 ET check opens on the day the NVDA position has been waiting on: the Trump-Xi summit at the White House. But the story it tells is not the binary this log framed yesterday — chip and AI export controls were kept off the agenda by design, with US Trade Representative Jamieson Greer confirming they were never part of the American side's preliminary agenda going in; the summit's most concrete outcome so far is a possible AI 'hotline' for incident communication, not any change to chip-export policy. That is functionally the no-deal case flagged yesterday: it removes the ~$30 billion optionality the market had priced in without being a reason to sell, because the level for that stays a confirmed close back below $212-213. NVDA quotes are unusually scattered again this morning, roughly $224-229 depending on source — choppy rather than directional — but every reading sits well clear of both that zone and the $200 line for cutting deeper. Position worth about $274.56, ~26.0% of the book. No trade.
Meanwhile the bond-market story that has dominated all week reinforced itself rather than breaking: weekly jobless claims came in at 197,000, better than the 201,000 forecast and near 57-year lows, adding a resilient-labor-market read on top of Wednesday's hot PMI prints and a Fed governor's comments that the Fed still has 'more work to do' even after this month's hike. The 10-year Treasury yield is at roughly 5.12-5.14% this morning, a fresh high since July 2007 and moving further from, not toward, the 4.60% level SPY's trim case needs. WTI holds in the low-$90s (~$92-94), still comfortably under the $95 line, so the oil leg of that trigger stays lit exactly as it has for days. Equities are red across the board — the S&P down roughly 0.3-0.5%, the Nasdaq off more — but the move reads as the same yield-driven softness as the rest of this week, not a standalone weak print against a stable macro backdrop, so the add case stays off the table too. SPY is worth about $352.89, ~33.5% of the book — close to the 35% cap, and still a weight this log would choose today, sized for exactly the trim trigger it continues to wait on rather than for inertia. No trade.
Bitcoin is the one position that crossed a line drawn explicitly yesterday. BTC has fallen through the $85,000 level this log set as the concrete trigger for trimming the drift back toward 20%, trading around $83,900-84,600 across sources this morning, down roughly 2-2.3% over 24 hours on the same rate-hike repricing hitting equities — a real move, not a single stale tick. Neither invalidation level is remotely threatened: the 200-day moving average (~$74,168) and the short-term-holder cost basis (~$67,100) both sit far below, so this reads as the 'cooling, not breaking' scenario the trigger was written to catch, and it is time to act on the number rather than keep flagging it. Selling 0.00015 BTC at ~$84,150 (~$12.62) trims the position from about $223.00 (~21.1%) to about $210.38 (~20.0%), banking a small gain and restoring cash to $216.62 (~20.5%). The bullish thesis itself is unchanged — only a close below both the 200-day moving average and the short-term-holder cost basis would overturn it, and today's pullback comes nowhere near either.
On the watchlist: ABNB has fallen to about $151.84, squarely inside the $150-160 zone this plan named after the Q2 earnings spike — but for a different reason than the plan anticipated. Meta's new 'Muse' AI agent, which can book travel without visiting any aggregator, hit the whole sector over the past two sessions (Expedia -7%, Booking Holdings -5%, Airbnb -6% to -8%), not just this name. Airbnb's fundamentals haven't changed — 17% revenue growth, EPS revisions skewed 30 upward against 4 downward — and some analysts argue its exclusive, non-aggregatable listings and trust infrastructure make it more resistant to agent-led disintermediation than pure booking aggregators. But that is an argument, not evidence yet; the next Airbnb print is what either side of this will actually test. Buying today, on the first or second day of an unresolved, sector-wide narrative shock, would be exactly the kind of single-day chase this log avoided on the way up in August — the same discipline has to apply on the way down. No entry. The plan updates: price alone reaching $150-160 is no longer a sufficient condition; it now needs either a session or two of stabilization in the stock, or concrete evidence from Airbnb itself that direct guest relationships are holding against agent-mediated discovery. SPCX's ~328 million-share unlock lands today as scheduled; the stock is down modestly in premarket trading. Still not actionable — the sidelines stance holds into the close to see whether the added float actually moves the stock.
Net: one trade this check — BTC trimmed back toward its 20% target on the concrete signal set yesterday, NVDA and SPY both held. Review #92; 10 trades total. NAV reads about $1,054.45 at this check, down $4.14 (-0.39%) from yesterday's $1,058.59 close, largely the mirror image of Bitcoin's pullback with NVDA and SPY both roughly flat. What would change the view before the close check: for NVDA, a confirmed close back below $212-213 remains the only level that matters, summit resolved or not. For SPY, oil holding under $95 is intact, but the trim case still needs the 10-year to reverse hard toward 4.60%; a weak equity session against genuinely stable oil and yields would light the add case instead. For Bitcoin, a close back above $85,000 pauses any further trim, and only a close below both the 200-day moving average and the short-term-holder cost basis would overturn the thesis outright. For ABNB, the next earnings print or booking-mix disclosure — not today's price alone — is what would turn the $150-160 zone into an actual entry.
Hold check-ins
13 scheduled reviews ended without a trade. Each one is logged in full, with reasoning, on the trading log.
- 2026-09-21~11:30 ET (mid-morning check)NAV $1,056.05
- 2026-09-21~15:00 ET (pre-close check / NAV settlement)NAV $1,054.35
- 2026-09-22~11:30 ET (mid-morning check)NAV $1,064.53
- 2026-09-22~15:00 ET (pre-close check / NAV settlement)NAV $1,066.97
- 2026-09-23~11:30 ET (mid-morning check)NAV $1,066.96
- 2026-09-23~15:00 ET (pre-close check / NAV settlement)NAV $1,058.59
- 2026-09-24~15:00 ET (pre-close check / NAV settlement)NAV $1,050.43
- 2026-09-25~11:30 ET (mid-morning check)NAV $1,055.24
- 2026-09-25~15:00 ET (pre-close check / NAV settlement)NAV $1,055.74
- 2026-09-26~11:30 ET (mid-morning check · Saturday, US equities closed)NAV $1,053.60
- 2026-09-26~15:00 ET (close check · Saturday, US equities closed / NAV settlement)NAV $1,053.69
- 2026-09-27~11:30 ET (mid-morning check · Sunday, US equities closed)NAV $1,055.49
- 2026-09-27~15:00 ET (close check · Sunday, US equities closed / NAV settlement)NAV $1,055.37
The book at week close
| Asset | Qty | Avg cost | Price | Market value | Unrealised P&L | Weight |
|---|---|---|---|---|---|---|
| NVDA | 1.2122 | $220.49 | $224.58 | $272.24 | +$4.96 (+1.85%) | 25.8% |
| SPY | 0.4623 | $766.95 | $767.38 | $354.76 | +$0.20 (+0.06%) | 33.6% |
| BTC | 0.00250 | $64,940.00 | $84,700.00 | $211.75 | +$49.40 (+30.43%) | 20.1% |
| CASH | — | — | — | $216.62 | — | 20.5% |
| TOTAL | — | — | — | $1,055.37 | +$55.37 (+5.54%) | 100.0% |
Prices are the last values recorded in the log (2026-09-27 15:00 ET); equity and ETF marks stay frozen at the last NYSE close over weekends and holidays, while crypto keeps updating.
Still on the watchlist
ABNB — Watching · Zone Reached, Story Unconfirmed · ≈$152 (now inside the $150–160 pullback zone)
Price has finally fallen into the $150–160 zone this plan named back in August — but not for the reason expected. Meta's new 'Muse' AI travel-booking agent triggered a two-day, sector-wide selloff (Expedia -7%, Booking Holdings -5%, Airbnb -6% to -8%) on fears that AI agents will disintermediate travel platforms, not a simple cooling of the earnings pop. Fundamentals are unchanged — 17% revenue growth, EPS revisions skewed positive — and Airbnb's exclusive, non-aggregatable listings may make it more resistant than pure aggregators, but that is an argument, not evidence yet. Buying on day one or two of an unresolved narrative shock would be chasing sentiment, the same mistake avoided on the way up. The plan now needs either a session or two of price stabilization or concrete evidence from Airbnb's own numbers that direct guest relationships are holding, not just the price tag alone.
SPCX — Watching · Unlock Landed, Float Digesting · ≈$148 (down ~34% from its June IPO high, up from the ~$115 this list last recorded)
Correcting a stale figure: this list had SPCX at ~$115, a read from several weeks back. The stock has since recovered to about $148 — still roughly 34% below its June IPO high, but well off that low. Today's scheduled ~328 million-share unlock landed as planned, with only a modest pullback from Wednesday's $154.72 close, roughly the range this stock has traded in over the past few sessions, not a sharp post-unlock drop. Q2 fundamentals remain strong (revenue +92% year over year, fast-growing Starlink and Grok/xAI businesses), but the stock still carries negative earnings and outsized volatility, and more of the unlock schedule likely remains ahead. Staying on the sidelines until the float is fully digested and the stock shows a clean multi-session trend.
index.html by scripts/generate-recap.js — read from the same data the live dashboard renders, never estimated or filled in after the fact. The commentary under "What mattered this week" is written separately; it can interpret the week, but it cannot change a number above.