Expression memo — It's 1997-98, not 1999: the melt-up has years left

Generated 2026-09-07 (born green at 72.3 on E288, EARLY flag at +0.0% excess vs SPY · window 2026-09-04 → 2029-09-04 · 4 mentions, 1 episode). Everything below the "spoken core" is DERIVED — OURS — NEVER SCORED. Verdicts and the scoreboard trace only to the spoken plays.

Read the EARLY flag honestly: it is an artifact, not a signal. The flag measures excess return since first mention, the idea was born on E288, so +0.0% is arithmetic, not opportunity. QQQ is already three years into the AI run. Nothing here says the entry is cheap — only that four voices independently say the top is not close.

Why this one went straight to green: it is the rare unanimous four-voice call, and each of them put a number on it. Sacks asked the question cold ("does this feel like 98 or 99 to you?") without teeing anyone up, and got four independent datings that all land pre-peak.

The spoken core (what IS scored)

• QQQ (primary) — large-cap tech beta, the melt-up index.
• ARKK (adjacent) — high-beta speculative growth; rips hardest late-cycle and breaks first at the top.

Spoken receipts, all E288: - Chamath, strength 3, the longest runway: "I think the thing is we're in the early part of the euphoria ... I don't see this stopping. We're at the beginning of the beginning. I think it's going to be like this for another probably three years." His mechanism is explicit and falsifiable: "bubbles burst when you realize that you priced way too aggressively forward, that instead of paying 50 times, you should have paid 15 times or five times." - Sacks, strength 2: "I'm starting to wonder if this is 1998 or 1999 ... I think it's more 1998 than 1999 because I don't think we've seen the peak yet." He also placed the whole table on the calendar: "You're 98, I'm 97, he's 96 We're further along than 96 for sure." - Jason, strength 2, the caveated bull: "I think it's 97, 98 ... I do think we are seeing at the late stage a disconnection in reality and valuations, where companies are being given far too much credit." He is pre-peak on the index and openly bearish on late-stage private marks — "unproven entrepreneurs getting a 50 or 100x is a little bit dangerous folks." - Friedberg, strength 2, the analytical spine: "With the.com boom, it was all like metrics that weren't dollars. What we're seeing now is revenue and profits and growth in revenue and profits that we've never seen before. ... it is all real dollars flowing versus speculative utilization." This is the leg that makes the thesis more than vibes — it names the thing that would have to break.

The arithmetic they did on air: Jason anchored the bust to "spring of 2000," so their datings convert to roughly 2.5 years (Sacks), 3 years (Jason and Chamath's own number), stretching to 4 in Jason's summary of Chamath. The scored window takes the middle at 36 months.

Second-order beneficiaries (derived)

• ai-compute-capex-supercycle (green, 85.7, MOVING) is the earnings substrate of this idea, not a duplicate of it. That idea says the capex keeps flowing; this one says the multiple on that capex keeps expanding. They can diverge — capex can run for 24 months while the multiple tops out first. Do not size them as one position; that is double-counting the same macro.
• enterprise-saas-moats, spacex-multiple-ways-to-win, nvidia-open-source-champion are individual lanes inside the melt-up and already own their tickers (IGV/MSFT/CRM, the SpaceX complex, NVDA). QQQ overlaps all of them. Net the overlap on /instruments before adding index exposure on top of names.
• The IPO window is the melt-up's distribution mechanism. Sacks on air: every San Francisco IPO in history summed is "only one quarter the size of the anthropic IPO," with OpenAI behind it. That makes anthropic-ipo (watch, 52.2) and ipo-boom-2026 (dormant, 6.7) derived beneficiaries — and it is the cleanest leading indicator here. A melt-up dies when the IPO window shuts.
• gambling-speculation-2026 (dormant, 14.1) is the purest euphoria beta we already own a lane for — retail speculation complex. If Chamath's "three more years" is right, that dormant idea is mispriced by our own board.
• The SF housing tell (unscored, no instrument). Sacks calls ultra-luxury SF to "5,000 a foot London type prices" off IPO wealth; Jason says he is clearing out all his SF property inside six months. Two full-weight voices on opposite sides of the same trade, and nothing on the board can score it. Watch it as a sentiment gauge, not a position.

Supplier / product-level drivers (derived — what to actually watch)

• Revenue quality is Friedberg's own stated test. His leg is "real dollars flowing versus speculative utilization." The kill signal is therefore specific: AI revenue that turns out to be circular (vendor financing, compute-for-equity, related-party bookings). Track the share of frontier-lab revenue coming from other frontier labs and from their own investors.
• Price-to-sales at the late-stage tier, which is Jason's number: he flagged "50 to 100 times top line revenue" as the unsustainable zone. When public comps print those multiples, his 97-98 becomes 99.
• The IPO calendar (Anthropic ~Oct 2026, OpenAI behind it). Successful mega-listings extend the runway; a broken one ends it.
• Secondary/cash-out volume — Sacks explicitly attributed the SF housing bid to "the secondaries, the small cash outs people have been able to do anticipating some of these events." Insider selling accelerating into the IPOs is the classic late-cycle marker, and he named it approvingly.

Options angles (derived)

• The window and the option chain do not line up, and that is the main execution problem. The idea evals 2029-09-04 (36 months), while listed LEAPS reach roughly Dec 2028. Any option expression here needs one roll, or it expires before the thesis is scored. Size for that.
• QQQ LEAPS calls (Jan-2028, rolling to Jan-2029) are the direct expression. QQQ IV is usually moderate even after a run, so outrights are defensible here rather than spreads.
• ARKK is where IV is rich — call spreads, not naked longs. ARKK is the correct instrument for the euphoria leg specifically, since Chamath's thesis is multiple expansion, and ARKK is nearly pure multiple.
• The patient expression, given the flag is an artifact: sell cash-secured puts on QQQ into any 10%+ drawdown instead of buying the top. Four voices saying "no peak yet" is an argument for buying dips, not for buying breakouts.
• Degen version: ARKK LEAPS calls, small. Note that this is levered to the leg most likely to break first even if the thesis is right — a 1998-99 analogue rips the index and still shreds high-beta on the way.

What kills it (invalidation — including the opposes already in the pool)

1. The scoreboard says these four have no demonstrated edge in equities. Current skill weights: Jason/equities 1.09, Chamath/equities 0.99, Sacks/equities 0.95, Friedberg/equities 0.90 — every one of them within a hair of coin-flip on 55, 85, 55 and ~50 resolved calls respectively. This green rests on unanimity, not on accuracy. Unanimity among four people who are all long the asset is the weakest kind of signal this system produces.
2. Their actions are more cautious than their dating. Jason spent the same episode telling founders to sell 10-20% of their holdings and saying he is "clearing everything out in the next six months" on his SF property. Chamath's own framing was "keep calm and carry on." Nobody at that table said they were adding risk.
3. us-debt-spiral-long-rates (green, 74.6, MOVING, +19.5% excess) is the direct multiple killer and it is the strongest-scoring bearish idea on the board. The 30Y at multi-decade highs is precisely what ends a multiple-expansion regime. These two greens are in tension; hold both only with that understood.
4. equity-lost-decade-post-zirp-2022 (dormant, 13.5) is the literal opposite thesis and is still live to 2032. It is dormant, not dead.
5. Chamath's own commoditization call undercuts the revenue leg. In the same episode he opposed frontier-ai-premium — capabilities "matched or close to matched ... within the next three or four months" and cost per unit of intelligence "driven further and further down." If he is right about that, Friedberg's "real dollars" compress, and the profits underneath the melt-up thin out. He is simultaneously the most bullish voice on duration and the most bearish on frontier pricing power.
6. The window is long enough to be right and still score a MISS. A top in 2027 leaves this idea looking correct for eighteen months and then closes MISS at 2029-09-04. Do not read a rising conviction score in 2027 as vindication.
7. The idea's own eval: 2029-09-04. Stale memo if the band drops — memos never refresh themselves.