Cymetica
Specialized AI
PAPER
Menu
Dark Mode
Plain English Mode
PAPER TRADING MODE — Enable real trading on your Account page
Back

AI Adaptation Research Report — CARG (CarGurus, Inc.) · BlackSwan constituent

AI Adaptation Research Report — CARG (CarGurus, Inc.) · BlackSwan constituent

EventTrader research desk · generated 2026-11-22 13:06 UTC · report ai-adapt-carg-20261122 · narrative: AI analyst, checked against the evidence pack

Company CARG — CarGurus, Inc. · SIC 7374 Services-Computer Processing & Data Preparation
Book AMF-BLACKSWAN (short) — BlackSwan
Composition algorithm Swan Composite v1.1.0 — 50% AI Adaptation Score + 25% AI Story Score + 25% market verdict, conviction-weighted
Anchor filing 10-Q filed 2026-11-05 (period 2026-09-30)
Adoption level 0 — absent · deterministic
AI Adaptation Score 9.4 / 100 · run 2026-11-22 · sector software percentile 1.4
Price 32.49 USD
Place in the book rank 2 of 10 · composite 28.5 · story 50.2 · market verdict 44.9 · weight 12.7%

1. Verdict

  • BlackSwan short book — the anchor filing contains zero AI sentences, so the score reads CARG as a non-adopter by its own disclosure [S5]. The one-line reason: a 9,681-word 10-Q that never substantively mentions artificial intelligence [S1].
  • Anchor filing: the CarGurus 10-Q filed 2026-11-05 for the period ending 2026-09-30, accession 0001193125-26-449875 [S1][S2].
  • Adoption level: "absent" (final level 0, lexical 0, no model read) [S5].
  • Score: AI Adaptation Score of 9.428, sitting at the 13.8th percentile of the software universe and the 1.4th percentile within its sector [S5].
  • What would move it: a later filing that substantively discusses AI — deployment, proprietary systems, or disclosed investment — which the score could then read and credit [S5].

2. The company

The pack identifies the filer as CarGurus, Inc., CIK 1494259, trading as CARG on Nasdaq [S4]. Its SEC classification is SIC 7374, "Services-Computer Processing & Data Preparation" [S4]. For scoring purposes the platform places CARG in the software sector, defined as "software companies (SIC 7370–7379)" [S5]. The filing reviewed is the company's quarterly report on Form 10-Q, filed 2026-11-05, covering the period ending 2026-09-30 [S1][S2]. Beyond identity, sector, and the anchor filing, the pack does not describe CarGurus' business, products, or operations, so none is asserted here.

3. What the filing says about AI

The pack records zero AI sentences in the anchor filing: ai_sentences is 0, substantive AI hits are 0, risk hits are 0, and the list of grounded quotes is empty [S5]. No filing sentence mentioning artificial intelligence was supplied in the evidence pack for this report. Accordingly, the filing reviewed does not state anything about AI that this report can quote — there is no verbatim sentence to cite, because the scored document carries none within the 9,681 words the model measured [S1][S5].

4. The AI Adaptation Score

The adoption level is "absent": a final level of 0, a lexical score of 0, and no model read (model_status is "none", confidence and credibility null) [S5]. Because no language model pass ran on this filing, there are no sub-scores from a model read to report — that layer simply did not engage, which is itself a consequence of the filing containing nothing about AI to read [S5].

The headline AI Adaptation Score is 9.428, with a margin of 23.759 [S5]. Its components, as given, are: level 0.0, greenfield 0.0, credibility 0.0, trajectory 5.958, and investment 3.47 [S5]. The entire positive score, in other words, comes from the trajectory and investment components; the level, greenfield, and credibility components are all zero, consistent with a filing that discloses no AI language [S5]. The platform assigns CARG to the "non_adopter" cluster within the software sector [S5].

On a relative basis, CARG lands at the 13.8th percentile of a 6,561-name universe and at the 1.4th percentile of its 590-company software sector [S5]. The sector reading is the more severe of the two: within software companies, CARG scores lower than roughly 98.6% of peers on this measure [S5]. These figures are reported as given and are not recomputed here.

5. Trajectory

The previous periodic report in the pack is the CarGurus 10-Q filed 2026-08-06, accession 0001193125-26-338067 [S3]. That prior filing was also scored at level 0, "absent," with 0 AI sentences and a substantive density of 0.0 [S5]. The level therefore did not move: CARG was absent then and is absent now, two consecutive quarterly filings with no substantive AI disclosure [S3][S5]. The trajectory component of 5.958 reflects the scoring model's own calculation and is reported, not reinterpreted [S5].

6. Recent developments

The pack's headline feed is empty: its status is "empty" with no items over the 45-day window [S5]. There are no recent developments to report from the evidence provided. For reference, the pack also carries a non-stale CARG quote of 32.49, supplied for context only [S6].

7. What would change the verdict

The verdict rests on a measurable fact — the anchor filing contains no AI sentences [S5]. To move CARG out of the BlackSwan book, a later filing or model read would have to show what this one does not: substantive AI disclosure the score can read. Concretely, that means a future 10-Q or 10-K that discusses deployed or proprietary AI systems, names AI investment (for example R&D or capex tied to AI), or describes AI-driven outcomes — language that would register as AI sentences, lift the lexical and level components above zero, and potentially trigger a model read with its own sub-scores [S5]. Until a scored filing carries such content, the components that are currently zero — level, greenfield, credibility — have nothing to credit [S5].

8. Risks and caveats

A filing-based score sees only what a company writes in the specific document scored. It cannot see AI work a company does but does not disclose in its 10-Q: internal tooling, product features described elsewhere, or engineering effort that never reaches the filing's text. A keyword-and-model read of one quarterly report is therefore a measure of disclosure, not of underlying capability — a company can build with AI and still score "absent" if its filing is silent. The score is also anchored to a single 10-Q [S1]; filings vary in how much narrative they carry, and a quarterly report is typically terser than an annual one. The verdict would be falsified if the scored filing in fact contains substantive AI language the pipeline missed, or if a later filing from CARG shows the disclosure described in section 7 and the level rises above "absent" [S5]. The headline feed being empty [S5] also means no outside events are reflected here.

9. Watch list

  • Next filing: the subsequent CarGurus periodic report; the current anchor covers the period ending 2026-09-30 and was filed 2026-11-05 [S1][S2].
  • Model read: none has run on this filing (model_status "none"); a future filing with AI content could prompt one [S5].
  • Next scoring run: this report reflects the 2026-11-22 run; the verdict should be re-checked on the following run against any new filing [S5].

Place in the book

  • Rank 2 of 10 in AMF-BLACKSWAN (Swan Composite v1.1.0, scoring run 2026-11-22): the book holds the lowest composites [S5].
  • Composite: 0.50 × 9.4 (AAS) + 0.25 × 50.2 (story) + 0.25 × 44.9 (market) = 28.5 [S5].
  • Market verdict 44.9: over the last 90 days CARG fell more and rose less than the software universe's median daily return (down-capture 0.89, up-capture 0.68, L = -0.20; 36 down / 18 up days) — the Failure-To-Adapt fund's selection signal, 50 = moves with the software universe's median daily return [S5].
  • AI Story Score 50.2: no AI story or a neutral one in the AI stories written about the company [S5].
  • Weight: rank factor 1.89 (rank 2 of 10) × story factor 1.00 × market factor 1.05, normalised across the book and clamped to 5–20% → 12.7% [S5].

Sources

Method and disclaimer

The AI Adaptation Score reads each company's own periodic SEC reports (10-K, 10-Q, 20-F, 40-F): deterministic features (AI sentences by section, proprietary-model, deployment, outcome, investment and vendor language, boilerplate and negation excluded) set an adoption level on a 0–4 rubric; a model read of the filing's AI sentences confirms the level and scores credibility and greenfield (the business built with AI from the ground up rather than retrofitted); the score combines level, greenfield, credibility, trajectory against the previous report and investment, measured against the company's sector, with greenfield carrying the highest reward. The books are composed by Swan Composite v1.1.0: each company's composite is 50% AI Adaptation Score + 25% AI Story Score + 25% market verdict — the AI Story Score is the sentiment of the AI stories written about it, the belief network of companies named in the same stories and the price reaction to those stories (no story scores neutral); the market verdict is the Failure-To-Adapt fund's selection signal, the signed down/up capture ratio of the stock against the software universe's median return over 90 days (a stock that falls more and rises less than its peers scores below 50). AMF-BLACKSWAN holds the ten lowest composites that pass the liquidity screen and the new-listing gate and AMF-WHITESWAN the ten highest, weighted by conviction (rank, story agreement and market agreement, each leg 5–20%), recomposed on every scoring run. Every quote in this report is the filing's text verbatim; every factual sentence names its source; a source the platform could not reach is reported as missing, never filled in.

This report is information, not investment advice. Figures come from the cited sources as of their dates. Nothing here is a recommendation to buy, sell or hold any asset.