Experiments
Labs — experiments on our own research
Every report we run, every concern we track, every move we explain accumulates into a data set nobody else has. Labs is where we publish quick experiments over it — cross-sections of our own research, not repackaged market data. Check back often: experiments rotate, get promoted into the product, or get retired as we learn. Every number here traces back to a report or a dated, sourced event.
Quality Universe Index — an equal-weight, hypothetical index over the whole quality universe, recorded nightly next to the S&P 500 from its inception close. It graduated to the Universe page.
Calibrate · did our signals actually help?
collecting · first edition ≈ May 2027Every six months, Calibrate will rank the quality universe best→worst by what the stocks actually did, and for the extremes publish the bridge: every dated signal OffHours gave on that stock — Buy Zone changes, concerns easing or worsening, setup lights, explained moves, valuation gaps, report verdicts — laid out in order next to the outcome. What we said, dated, next to what happened. Both directions, every time.
Signals led
A favorable signal (for a winner) or an unfavorable one (for a loser) came before most of the move. The best we can honestly claim.
Signals lagged
The signals eventually agreed with the outcome — but only after most of the move had already happened. Right, late.
Signals were silent
OffHours had little or nothing dated to say in the window. The outcome is the market's story, not ours.
Signals misled
The signals mostly pointed the wrong way — including any setup-complete on a bottom-quintile stock. This bucket is the reason to trust the other three.
The humility line, up front: ~185 stocks, 6 months — luck dominates single outcomes; patterns emerge over cycles. A bucket is a label on a dated timeline you can read yourself, never a score. And this page will never be marketing: the misled bridges ship in every edition, with the same layout as the wins.
What's recording, since when
| Signal | Recording since | Events | State |
|---|---|---|---|
| zone Buy Zone transitions | 2026-08-14 | 560 | Grade changes have a permanent per-ticker log; since 2026-09-04 the nightly gate also writes the tenant-scoped calibrate stream. |
| fear Fear ledger (named, eased, worsened) | 2023-09-17 | 518 | Every fear is dated and sourced; eased/worsened events come from the daily scan. |
| setup Setup-complete flips | 2026-09-07 | 56 | Started recording 2026-09-04 with this build — the nightly gate now captures every on/off flip. No history exists before that; the first edition will say so. |
| alert Moves explained (±3% days) | 2026-08-13 | 115 | The shared per-ticker journal of the big-move days we explained — the same moments the move alerts fired on. |
| gap Reverse-DCF gap snapshots | 2026-09-07 | 7799 | Daily implied-vs-delivered growth points on 184 tickers; the timeline shows inflections, not every day. |
| report Dive-deep reports (dates + verdicts) | 2026-05-06 | 49 | From the shared research archive — dates and verdicts only, never anyone's personal purchase data. |
First narrated edition ≈ May 2027 — six months after the youngest core stream started recording. Returns are price returns (dividends excluded), applied the same way to every stock.
Assembler preview — no narration, no ranking
NVDA · last ~6 months · price -41.7%- 2026-07-31reportDeep-dive report published — verdict: Attractive Valuation
- 2026-08-06fearA fear eased — Amazon guided to $220B and Alphabet raised guidance to $195-205B in 2026 capex; Wall St. consensus for top-5 hyperscaler 2026 capex now $733B, up over 100% from June estimate
- 2026-08-09fearFear named — Hyperscaler Capex Sustainability and AI ROI Skepticism — The bull case depends entirely on hyperscalers (Microsoft, Alphabet, Amazon, Meta) continuing $300B+ in 2026 AI infrastructure spending; any pullback in capex guidance if enterprise AI fails to show adequate ROI would hit NVIDIA's revenue directly and immediately. This is framed as resolving within the next 4-6 quarters as capex/revenue ratios either prove out ROI or correct, making it a time-limited overhang rather than a permanent impairment, especially as Alphabet just raised full-year Capex guidance to $195-205 billion.
- 2026-08-09fearFear named — Stock Underperformance Despite Record Fundamentals ('The Disconnect') — NVDA shares fell from a May 14, 2026 record of $235.47 to as low as $192.53 by June 26 (down ~18%) even as fiscal Q1 FY2027 revenue hit a record $81.6 billion, up 85% year-over-year, and the company has beaten EPS estimates in the three most recent quarters shown in earnings data. This is a valuation/sentiment reset rather than a deterioration in the underlying business, since guidance and demand from hyperscalers reportedly still outstrip supply.
- 2026-08-09fearFear named — Insider Selling by Executive Mark Stevens — Director Mark A. Stevens sold a combined 885,000 shares on June 18, 2026 at prices between $209.70-$210.44, above the current $198 level. While notable in size, single-director sales at a company with NVIDIA's market cap are common portfolio diversification and not indicative of a structural business problem.
- 2026-08-09fearFear named — China Export Restrictions and Loss of China Data Center Revenue — NVIDIA has explicitly excluded China data center revenue from guidance, and BIS shifted to a case-by-case review with 25% tariffs and volume caps on chips like the H200 bound for China, alongside the emergence of competitive open-source Chinese models like Moonshot AI's Kimi K3. This is a real revenue headwind but is time-boxed to the current policy regime and could resolve or worsen with U.S.-China trade negotiations rather than being a permanent structural loss of the moat itself.
- 2026-08-09fearFear named — Circular Financing Concerns Among AI Ecosystem Players — Renewed fears about circular financing arrangements—where NVIDIA, OpenAI, Oracle, and CoreWeave have interlocking vendor equity stakes and take-or-pay compute deals—triggered a nearly 5% single-day drop on July 27, 2026, as investors worried end-demand is being artificially inflated. This is time-limited in that it resolves once hyperscaler cash-pay revenue is verified through actual earnings, but it can recur each quarter until proven out.
- 2026-08-09fearFear named — DOJ/FTC Antitrust Probe into Software-Hardware Bundling — Reports of a coordinated DOJ and FTC investigation into NVIDIA's market dominance and potential anti-competitive bundling of its CUDA software with hardware represent a genuine legal risk, but a ban on bundling (if it even occurred) would be a headwind to margins/lock-in rather than a business-model-ending event given NVIDIA's technology lead persists independent of bundling practices.
- 2026-08-09fearFear named — Rising Treasury Yields Compressing Tech Valuations — Broad rotation out of high-beta technology names as rising treasury yields compress valuations across the semiconductor sector, unrelated to NVIDIA's specific execution. This is a market-wide multiple de-rating rather than a demand problem for NVIDIA's chips.
- 2026-08-09fearFear named — Fed Rate Path and Macro AI Infrastructure ROI Anxiety — A broad market narrative questioning whether trillions in AI infrastructure spend will generate adequate ROI has hit chip-adjacent names (Applied Materials, Arm, TSMC) simultaneously, driven partly by Fed signals of prolonged high rates rather than NVIDIA-specific news.
- 2026-08-10fearA fear eased — CoreWeave closed a $2.6 billion loan facility on Aug 10, 2026 backed by a diverse group of AI, financial-services and tech customers, with lenders citing confidence in long-term GPU demand
- 2026-08-13fearA fear eased — Analysis published Aug 13 shows Wall Street has historically underestimated hyperscaler capex growth, with 2026 consensus now $733B vs $361B originally forecast
- 2026-08-13alertMove explained — +3.0% — Nvidia's rise appears tied to continued analyst optimism, including a Goldman Sachs ana…
- 2026-08-18fearA fear worsened — Nvidia expands OpenAI-linked financing web via ~$1.5B (closing in on $3B) SB Energy stake for OpenAI's Ohio data center, while Fortune reports the OpenAI-Nvidia data center deal size came in $145B lower than earlier reported, reviving 'arti
- 2026-08-19fearA fear worsened — Chinese AI firms reportedly accessing Nvidia GB300 compute via Southeast Asian data centers, spurring push for Remote Access Security Act to close the loophole
- 2026-08-19fearA fear eased — Nvidia begins actual H200 chip shipments to China, with ByteDance and Tencent each receiving ~10,000 units under case-by-case licenses
- 2026-08-21fearA fear worsened — Chip stocks fell this week as the 30-year Treasury yield hit a 19-year high, not on chip-specific news
- 2026-08-26fearA fear eased — NVDA Q2 FY2027 revenue hits $96.2B (+106% YoY) and guides to ~70% revenue growth in FY2028, above the ~45% Street estimate
- 2026-08-26fearA fear eased — AWS commits to buy 2 million Nvidia GPUs and adopt new Vera CPU; Nvidia CFO guides top-5 hyperscaler 2026 capex to $1.3T (from $800B) and Q3 revenue to $108B vs $104.2B street estimate
- 2026-08-27alertMove explained — +8.7% — Nvidia jumped after reporting blowout quarterly earnings and stronger-than-expected…
- 2026-08-28fearA fear worsened — Nvidia's reported $12.9B Hugging Face acquisition must clear mandatory DOJ/FTC antitrust review, unlike its Groq and Poolside deals which used license-plus-equity structures to avoid HSR filing
- 2026-08-28alertMove explained — −4.6% — Nvidia shares fell after the chipmaker paused parts of a financing program for AI cloud…
- 2026-09-01fearA fear worsened — Oracle borrows $43B to cover negative free cash flow and falls 5% on Sept 1; SB Energy IPO filing shows $3.2B loss on just $139M revenue
- 2026-09-02fearA fear worsened — Director Mark Stevens files to sell up to 5 million more NVDA shares (~$1.09B) via Form 144 on Sept 2, pushing his total disposals since June toward $1.5B
+46 more dated events in the window not shown here.
preview label: signals led — A unfavorable-direction signal on 2026-08-18 predates the halfway point of the move (2026-11-10) — more than half the 41.7% decline came after it.
Labs is part of Pro. Upgrade to see the full experiments — the headlines below are free.
№ 1 · Where is the market wrong about growth?
new experimentFor every stock we've analyzed, a reverse DCF (with that stock's own calibrated inputs) reveals the growth today's price demands — compared against the growth actually delivered. Negative gap = the market prices in LESS than the record: pessimism. Positive = the price needs the company to beat its own history.
Median across 20 analyzed stocks: the market demands -5.0pp less growth than these companies have delivered.
№ 2 · How much growth runway is left?
sample growingEvery fresh deep-dive now judges how much room a company's growth story has left (market penetration, TAM saturation). The small number on each chip is experiment № 1's gap — how many percentage points MORE (+) or LESS (−) annual growth the current price demands versus what the company has delivered. The danger quadrant is LIMITED runway with a positive gap: the market demanding above-record growth from a company running out of room to grow. That combination is highlighted in red.
Across 13 analyzed companies: 2 long runway, 8 adequate, 3 limited. Pro sees the per-company breakdown and each one's price-vs-record gap.
№ 3 · Which sectors are cheap against their own history?
new experimentEvery stock in the universe is graded against its OWN 5-year valuation history — so grouping by sector dodges the usual apples-to-oranges "sector P/E" problem. A sector where most quality names sit in their Buy Zones is a sector the market has marked down wholesale.
| Sector | In a Buy Zone | Very Attractive | Share in zone |
|---|---|---|---|
| Financial Services | 7 / 10 | 1 | |
| Basic Materials | 7 / 13 | 1 | |
| Industrials | 26 / 50 | 7 | |
| Technology | 20 / 39 | 10 | |
| Consumer Defensive | 6 / 13 | 3 | |
| Consumer Cyclical | 10 / 28 | 4 | |
| Communication Services | 2 / 6 | 1 | |
| Healthcare | 6 / 24 | 1 | |
| Energy | 0 / 2 | 0 |
Recomputed every few hours · Nov 21, 12:01 AM ET
№ 4 · What is the market afraid of right now?
sample growingWe track the concerns that keep each cheap stock cheap and grade every sourced, dated movement. Aggregated over the last 30 days, they form a bottom-up concern index — built from receipts, not surveys. It fills in as the concern scans accumulate.
№ 5 · Sentiment month or fundamentals month?
sample growingEvery ±3% move on a tracked stock gets classified: market-wide (the index explains it) or idiosyncratic (its own story). The split tells you whether to read the tape or ignore it.
20% of the 20 big moves in the last 30 days were market-wide rather than company news — a fundamentals stretch.
amber = market-wide · green = company stories
№ 6 · Do quality-stock dips recover?
sample growingEvery ±3% dip we journal gets graded 30 days later: higher or lower? Over time this becomes an honest scoreboard for the oldest question in quality investing — whether the scary red days in proven businesses were opportunities.
Of 24 graded dips, 38% were higher 30 days later.
Not a promise — a running score, updated as every dip ages past 30 days.