Insurance & reinsurance — cluster pass, 2026-09-22

Insurance & reinsurance — cluster pass, 2026-09-22

Slot cluster-insurance. Drafted by Claude in an unattended cloud routine. Nothing here is human-verified. Every figure carries its source; anything from background knowledge is marked [background]. Search-result evidence is marked [search-summary] — evidence of the summary kind, not a primary document, exactly as the skill's rung 6 requires.

Eight members: ACGL, KNSL, RYAN, ALL, AIG, LNC, FAF, AEG. All eight are SEC filers (seven domestic, AEG a foreign private issuer), so rung 1 answered nearly everything and rungs 2–4 were not needed. Zero Equibles calls.

Stages carried by this pass: S0 (universe & fit, standard four-part test — no member sits in a japan_* cluster), S4 (industry & supply map), S6 (scuttlebutt), S7 (thesis + framework test). Per-company stages (S1, S2, S3, S5, S8, S9) stay with the name passes names-insurance-1/2/3.

The product test and the expert call stay ⏳ OPEN (user); S7 conviction stays ⏳ OPEN (user).

---

Variant perception

The slot note frames the falsifier correctly and then the cluster file has to say which side of it the evidence lands on: "pricing is cyclical and the edge accrues to the best underwriters, not the sector."

The evidence says the cycle has already turned, and that the reported numbers are still hiding it. Not forecast — visible in the segment tables of a filing from 2026-07-28.

These eight are not one sector. They are four machines with four different engines, and the watchlist's single insurance tag conceals that the variable driving each one is different:

The interesting question is not whether the soft market hurts. It is that the group's reported combined ratios are improving while its underlying combined ratios deteriorate, and the gap is being filled by releases of reserves set in the hard market. That is a filing-level fact in the one member that discloses it most cleanly, and it is the single most transferable thing this pass found.

---

§1 — The cycle, from a filing rather than from commentary

Arch's 2026 second-quarter release breaks out the reported combined ratio and the combined ratio excluding catastrophes and prior-year development. The two move in opposite directions.

Reinsurance segment, three months ended June 30 (ex-991release63026.htm, "Reinsurance Segment" table, 8-K filed 2026-07-28, accession 0000947484-26-000118):

The release states the cause in its own words: net premiums written fell "due, in part, to non-renewals, share reductions as well as targeted increased retrocessions." Gross premium is flat; the company is keeping less of it.

The same shape at the group level: combined ratio ex-cat and ex-development of 82.5% against 80.9% in Q2 2025. And in the insurance segment, gross premiums written −2.9%, net −5.1%, net earned −4.5%, with the ex-cat ex-development ratio at 91.6% against 90.6%.

The mortgage segment is the one to watch and nobody is watching it. Its loss ratio was 6.5% in Q2 2026 against −1.2% in Q2 2025 — it had been running below zero, reserve releases exceeding incurred losses. Combined ratio 22.8% against 15.2%, a 7.6-point deterioration, and underwriting income fell 7.6% on premiums that grew. Mortgage credit normalising removes an earnings source that required no premium growth at all.

Why this is the cluster's spine: Arch is the best-capitalised, most disciplined underwriter of the eight, and it is the one that discloses the split. If its underlying margin is deteriorating 2.7 points while it shrinks premium deliberately, the members that are not shrinking are taking the business Arch declined. That is the mechanism by which a soft market transfers losses from the disciplined to the eager, and it is happening now, not prospectively.

---

§2 — S4: Industry & supply map ✅

One value chain for the cluster, with each member placed on it.

  capital  ──▶  REINSURER  ──▶  PRIMARY CARRIER  ──▶  WHOLESALE BROKER  ──▶  retail agent ──▶  policyholder
(investors,        ACGL(Re)        ACGL(Ins), KNSL,          RYAN              (not held)
 ILS, debt)                        AIG, ALL, FAF,
                                   LNC, AEG

Read the chain the way the 2026-09-20 AI-capex file read its own: by asking who absorbs an outcome and who merely takes a fee on the way past.

Where the pricing power actually sits, and the trap in the broker case

The broker is the only link that takes no underwriting risk, which is the standard reason to prefer it through a soft market. The board carries that exact argument, from a sell-side P&C analyst on 2026-09-07:

"re-insurance broker is definitely my top subgroup, without a doubt." — Ryan Tunis, candidates/eisman-playbook-2026-09-07-p-c-stocks-worth-owning-the-ai-hedge-with-04-insurance-brokers.yaml (turn at 35:49)

The filings say the trap is that a percentage of a falling number is a falling number. Ryan Specialty cut its 2026 organic revenue growth guidance from high- to mid-single digits, with Q2 organic growth expected near zero, and the shares fell ~20% on the miss; Goldman Sachs downgraded to Neutral with a $35 target, from $42, citing a prolonged period of commercial P&C pricing moderation [search-summary]. The cause named in the trade press is the same one in Arch's filing: declining property rates and intensifying competition [search-summary].

The broker does not escape the cycle. It escapes the loss ratio and keeps the volume exposure, which in a soft market is the half that is moving.

Customer concentration — the analogue, and why it is absent

The 2026-09-20 AI-capex pass located concentration at every link. Insurance has no customer concentration of that kind, and pretending otherwise would be the error. The structural analogue is counterparty concentration in the other direction — who the carrier has ceded its risk to. Two readings, from XBRL (ReinsuranceRecoverables, latest tagged annual):

• LNC: $28.0bn of reinsurance recoverables at FY2025, against $11.3bn of total stockholders' equity at 2026-06-30. A life insurer that has reinsured its back book is 2.5x levered to its reinsurers' solvency, not to its own. That is the concentration risk in this cluster, and it sits in the life names rather than the P&C ones.
• ACGL: $4.3bn at FY2019, the most recent year the concept is tagged — the series stops, which is a disclosure gap, not evidence of a small number.

Gate: met. The chain is written end to end with pricing power located on it, sourced from the issuers' own filings at the reinsurance, insurance and mortgage links.

---

§3 — S0: Universe & fit, per member

Standard four-part surplus test (no japan_* member). Owner checks — circle of competence and the exclusion list — are dropped per ledger/research/README.md and WATCHLIST_PLAN.md §16. Each member gets the four-part score with named evidence, the surplus question, the falsifier, and the liquidity reading.

ADV is measured identically for all eight: mean of close × volume over the 21 sessions 2026-08-21 → 2026-09-21 (22 for ACGL), ROIC.ai daily bars. Prices are the 2026-09-21 close.

A note the test itself forces

The four-part test was written for technology. Applied to insurance it scores "liability absorption" as a pass for seven of eight members, because absorbing liability is literally the product they sell. That is not a moat finding; it is the test detecting the industry. The axis that discriminates inside this cluster is the fourth — a bottleneck that cannot be manufactured — and the honest answer for most of these names is that underwriting capacity is exactly the thing capital manufactures, which is why the cycle exists at all. The board says the same thing from the outside:

"we've seen this industry willing to accept returns that are unacceptable." — Ryan Tunis, transcript src_eisman-playbook-2026-09-07-p-c-stocks-worth-owning-the-ai-hedge-with.txt, line 563 (34:38 turn)

So a low score here means "this is a cyclical business", not "this is a bad company". The scores below should be read as where the name sits relative to the cycle, and S7 does the rest.

---

ACGL — Arch Capital Group ⚠️ 2/4

Surplus question — who keeps it? Arch does, when there is any: FY2025 net income $4,399m on $24.2bn of equity, and Q2 2026 annualised ROE of 18.0% (15.3% on an operating basis). But the company is handing the surplus back rather than deploying it — $1.2bn of buybacks in Q2 2026 alone, and diluted shares from 376m (FY2025 weighted average) to 341.2m on the 2026-07-30 cover page, ~9% in half a year. Management's capital allocation is itself the cycle call.

Falsifier. The ex-cat ex-development combined ratio. It went 77.2% → 79.9% in reinsurance and 90.6% → 91.6% in insurance year on year. Two more years of that arithmetic, with prior-year development normalising toward zero and the mortgage loss ratio already through zero (−1.2% → 6.5%), and the reported combined ratio crosses 100 without a single catastrophe.

Liquidity. ADV $341m/day; BVPS $68.04 at 2026-06-30 against $95.78, P/B 1.41. No constraint at any plausible size.

Verdict: ⚠️ 2/4 — informative rather than disqualifying. Arch fails the bottleneck test because the industry has no bottleneck; it passes everything a cyclical underwriter can pass. Hold it as the cluster's control: it is the disciplined actor, and the file uses its disclosure as the measuring stick for the others.

---

KNSL — Kinsale Capital Group ✅ 3/4

Surplus question — who keeps it? Shareholders, undiluted. FY2025 net income $504m against $88m in 2020 — 5.7x on a constant share count. This is the only member of the eight whose earnings growth and per-share growth are the same number.

Falsifier — and it is the sharpest one in the cluster. Kinsale has never underwritten through a soft E&S market. Its whole record (2020–2025) is a hardening one. The board carries the specific bear case, and it is about claims practice rather than pricing:

"which I think could be a problematic company" — Steve Eisman, candidates/eisman-playbook-2026-09-07-p-c-stocks-worth-owning-the-ai-hedge-with-03-knsl.yaml (45:54)

The claim file's own thesis field — the extractor's wording, not a spoken quote — records the mechanism behind that line: a relayed concern that Kinsale's small-account model echoes "an old AIG playbook of writing policies for small companies and then contesting claims", with Tunis pushing back that the loss ratios do not support the read. Both halves matter — the loss ratio is the falsifier and it is currently clean. A soft market is when a claims-handling story would show up in it.

Liquidity. ADV $70m/day on 22.8m shares — the thinnest of the eight after AEG and FAF, and the one where S8 sizing must respect it. P/B 4.01 ($89.25 BVPS), P/E 16.2 on FY2025 — by some distance the most expensive name here.

Verdict: ✅ 3/4. The only member that passes on something other than being an insurer.

---

RYAN — Ryan Specialty Holdings ❌ 1/4

Surplus question — who keeps it? Not the public shareholder, and this is the finding. From the filings:

Revenue more than doubled and the earnings attributable to the listed entity were lower in 2025 than in 2021. Of $214m of consolidated 2025 profit, $151m — 71% — was attributed to the noncontrolling interests of the Up-C structure. The 10-Q cover page for the quarter ended 2026-06-30 reads: "On July 27, 2026, the Registrant had 255,814,785 shares of common stock outstanding, consisting of 122,077,702 shares of Class A com[mon]" — so Class A is 47.7% of the shares and received 29% of the profit.

Against that, goodwill of $3,225m and long-term debt of $3,346m (FY2025 XBRL) sit on $509m of total stockholders' equity at 2026-06-30. The roll-up is financed at the listed entity; the economics are shared with the unitholders.

Falsifier. Organic growth. The whole case is that brokerage compounds through the cycle; Q2 2026 organic near zero is that case failing its first test in this cycle, and the Goldman downgrade names the reason as pricing rather than execution [search-summary].

Liquidity. ADV $74m/day. Market capitalisation on all 255.8m shares ≈ $10.2bn at $40.06. P/B and P/E are not meaningful here — 20.1x and 163x on the parent-only figures — and any screen that computes them on the listed entity's book will misrank this name badly. Use consolidated pre-tax income ($293m FY2025) against enterprise value instead.

Verdict: ❌ 1/4. The fee model is real and the channel position is the best in the cluster. The failure is that the listed security is a minority claim on it, levered, and priced on consolidated numbers it does not fully own.

---

ALL — The Allstate Corporation ❌ 1/4

Surplus question — who keeps it? Right now, Allstate, spectacularly — and that is the problem. XBRL net income, FY2020 → FY2025: 5,576 / 1,614 / −1,289 / −188 / 4,667 / 10,282. The 2025 figure is the best of the six years by more than 2x and follows two loss-making years. Revenues grew from $41.9bn to $67.7bn across the same span.

Peak earnings and peak analyst negativity are the same date. From the board, 2026-09-07:

"the personal lines subgroup, which I've not been more negative on in my career" — Ryan Tunis, candidates/eisman-playbook-2026-09-07-p-c-stocks-worth-owning-the-ai-hedge-with-01-pc-personal-lines.yaml (13:57)

The claim row records the arithmetic behind it: auto insurers raised rates ~60% when claims inflation required ~40%, and are now cutting — expected to persist through the end of 2027 (the row carries normalized_end_date: 2027-12-31).

Falsifier — structural, not cyclical, and it is the one that matters. The same row records that ADAS features are in roughly 40% of new vehicle sales against 15% three or four years ago, collapsing collision frequency. A soft market is a price problem and mean-reverts. A shrinking exposure base does not. Allstate's $10.3bn of 2025 earnings are being capitalised at 6.0x by a market that may be right to.

Liquidity. ADV $422m/day — the deepest of the eight. P/B 1.82, P/E 6.0 on FY2025 earnings. Caveat: Allstate's preferred stock is not tagged at a current date in XBRL, so the P/B is computed on total stockholders' equity and flatters the common slightly.

Verdict: ❌ 1/4. The cheapest-looking name in the cluster on trailing earnings, at the point in the cycle where trailing earnings are least informative.

---

AIG — American International Group ⚠️ 2/4

Surplus question — who keeps it? Shareholders, by subtraction rather than growth. This is the clearest capital-return story of the eight, and the XBRL makes the scale visible:

Revenue −39%, share count −34%, long-term debt −76%. The company shrank itself deliberately and bought in a third of its equity doing it — Tunis's claim row attributes the simplification to the divestiture of the last Corebridge stake, leaving a pure-play P&C name. The loss ratio improved from 87.0% to 59.6% over the same six years (XBRL, PolicyholderBenefitsAndClaimsIncurredNet ÷ PremiumsEarnedNet).

The board carries both the recommendation and a disclosed position:

"I am recommending AIG." — Ryan Tunis, ...-03-aig.yaml (18:24) "And full disclosure, everybody, I own AIG." — Steve Eisman, ...-01-aig.yaml (18:26), row records a recently-upgraded long

Falsifier. The self-help levers are finite. Shares are down a third and debt down three quarters — those are one-time repairs, and the residual business is the one most exposed to softening large-account pricing. When the expense and reinsurance synergies are banked, what is left is a price-taker in the worst-priced segment.

Liquidity. ADV $310m/day. P/B 0.97 ($76.70 BVPS) — the only P&C member below book. P/E 12.6 on FY2025.

Verdict: ⚠️ 2/4. Scores like Arch and is a different bet: Arch is a good underwriter facing a bad price, AIG is an average underwriter facing a bad price with a shrinking share count and a balance sheet already repaired.

---

LNC — Lincoln National ❌ 1/4

Surplus question — who keeps it? Unclear, and the balance sheet is why. Equity went $22,699m (FY2020) → $5,102m (FY2022) → $10,906m (FY2025) — a 78% drawdown and a partial recovery, on XBRL. Net income over the same span: 499 / 3,778 / 1,358 / −752 / 3,275 / 1,177. Neither series supports an estimate of normal earnings power; the equity move is a rate/AOCI artefact and the income series is dominated by assumption reviews [background].

Falsifier — and it is a counterparty question, not a market one. ReinsuranceRecoverables of $28.0bn at FY2025 against $11.3bn of equity at 2026-06-30. Lincoln has reinsured the back book, most recently a ~$5.8bn GUL transaction [search-summary]. That converts an actuarial risk into a credit risk on a small number of reinsurers, and the exposure is 2.5x equity. The reported RBC of above 420% at Q2 2026 [search-summary] is the right ratio to watch and it does not measure this.

Liquidity. ADV $87m/day. P/B 0.70 ($59.29 BVPS) — the cheapest book in the cluster, which is the market's verdict on exactly the above. P/E 6.8 on an FY2025 number nobody should annualise. Caveat: LNC's preferred is not tagged at a current date, so the P/B flatters the common. Morgan Stanley is at Equal Weight, $42 target against a $41.68 close [search-summary] — i.e. the most engaged outside estimate sees no return.

Verdict: ❌ 1/4.

---

FAF — First American Financial ⚠️ 2/4

Surplus question — who keeps it? The mortgage market decides, not the company. XBRL net income: 696 / 1,241 / 263 / 217 / 131 / 622 (FY2020 → FY2025). A 9.5x peak-to-trough swing with no change in competitive position — that is a volume business wearing an insurance licence.

Falsifier. Two, pulling opposite ways. Management expects muted residential purchase trends to persist until housing activity rebounds, while guiding commercial title to a possible record year on data-centre and energy transactions [search-summary] — which quietly makes FAF a second-order AI-capex name, a link the watchlist's cluster tags do not carry. Against that, the same coverage flags AI and regulatory pressure on the title model, including proposals to cut title-insurance requirements [search-summary]; the company is publicly opposing them [search-summary]. A regulatory change to the requirement is the kill, not a competitor.

Liquidity. ADV $59m/day — the thinnest of the eight. P/B 1.26 ($55.05 BVPS), P/E 11.4 on the recovered FY2025 number. Consensus targets $75–78 against $69.26 [search-summary].

Verdict: ⚠️ 2/4, and the most mis-clustered name here: its driver is housing and, increasingly, data-centre transaction volume, not the P&C cycle the slot was built around.

---

AEG — Aegon Ltd. ⚠️ 2/4

Foreign private issuer, IFRS. Read from the Form 425 filed 2026-08-20 (accession 0001193125-26-357719, "Interim Financial Information for the six-month period ended June 30, 2026") and the IFRS XBRL in companyfacts.

Surplus question — who keeps it? Shareholders, through shrinkage, and the filing is explicit: the interim dividend is 21 eurocents, "up 11% versus the prior year period", and the buyback was increased "by EUR 150 million to EUR 350 million". The weighted average common share count fell from 1,573m (1H 2025) to 1,489m (1H 2026), ~5% in a year. Total assets have gone €459.6bn (Jan 2022) → €317.2bn (FY2025) and insurance revenue €11,251m → €9,097m over the same span (IFRS XBRL). This is a deliberately smaller company returning capital.

Falsifier — and it is in the filing, not in the commentary. The capital-highlights table reads:

Own funds fell 4% while the capital requirement rose 4%, and the group solvency ratio dropped 15 points in six months — during which the buyback was increased by €150m. Outside commentary described the same period as a "capital beat" supporting a bigger buyback [search-summary]; that is true of the US RBC ratio and of operating capital generation, and not true of the group ratio. Both are in the filing. The difference between those two readings is the whole AEG case, and it is resolvable only from the primary document.

The filing also confirms the structural event: "progressed at pace with our planned relocation to the US", with Aegon having agreed to adopt the Bermuda solvency framework after a transition period (note 16). A redomiciliation changes the regulator that computes the ratio above.

Liquidity. ADV $40m/day on the NYSE ADR — the thinnest of the eight. Shareholders' equity EUR 4.94 per share at 2026-06-30, = $5.67 at EURUSD 1.14673 (ROIC, 2026-09-21), against a $9.09 close: P/B 1.60. That ratio understates the asset: IFRS 17 equity excludes the contractual service margin, so a life insurer's book is not comparable to a P&C insurer's book, and the cross-section table below should not be read across that line.

Verdict: ⚠️ 2/4.

---

§4 — S6: Scuttlebutt ✅ 8/8

Gate: one piece of evidence per member that did not come from the company. Met for 8 of 8.

The board is the cheapest scuttlebutt in the building — and this time it is on-topic

ledger/pipeline/gold/candidates/ holds a 2026-09-07 episode of The Real Eisman Playbook whose entire subject is this cluster: a sell-side P&C analyst taken through personal lines, commercial lines, reinsurance, brokers and specialty, subgroup by subgroup. Six claim rows were extracted from it. Every quote below was copied from the claim file or located by exact substring in the committed transcript, never typed.

Transcript note, recorded so a later reader does not chase it: the reinsurer list at line 556 is an auto-transcript mangling — "Swissery" and "Berenary" are not resolvable entities and were not turned into nodes or subject ids. Only "Arch" was used, and only because it matches an existing registry id.

Thin, and named as such: LNC, FAF and AEG carry search-summary evidence only — no primary outside document and no independent operational reading (no Glassdoor, no Blind, no channel checks). ACGL and RYAN have one outside data point each beyond the podcast. Only ALL, AIG and KNSL have scuttlebutt that is both named and on the record. The product test and the expert call stay ⏳ OPEN (user).

---

§5 — S7: Thesis + framework test ✅

Framework used: the lens table from _clusters/ai-capex-2026-09-20.md, adapted — the account skills baker-thesis and sacerdote-thesis are not available in routine runs, and a plain S-curve/adoption read is the wrong instrument for a capacity cycle. The adaptation is one substitution: where that table says "shortage", read "underwriting capacity", and the seller/buyer question becomes who is selling capacity into a glut and who is being paid a fee on top of it.

Cluster thesis

The hard market is over and the disclosure lags the fact. Reported combined ratios across the group are being held up by releases of reserves established in 2022–2024. The underlying ratios are already deteriorating — 2.7 points year on year in the cleanest disclosure available (Arch reinsurance), 1.0 point in insurance, 7.6 points in mortgage. The disciplined response is to write less, which Arch is doing at −10.4% net premiums written, and somebody else is writing that business.

Therefore: the slot's own falsifier resolves in favour of the underwriter, not the sector. There is no sector trade here. There are three or four individually sound positions and four names whose apparent cheapness is a function of where trailing earnings sit in the cycle.

The lens table

The cross-section

Price at 2026-09-21 close. Book is common stockholders' equity at 2026-06-30 (10-Q), earnings are FY2025 (10-K). ADV is 21 sessions to 2026-09-21 (22 for ACGL).

¹ Company-stated BVPS from the Q2 2026 release; the XBRL computation returns $68.00, which is how the method used for the other seven was validated. ² Up-C structure. Market cap is all 255.8m shares; earnings are the $63m attributable to the listed entity out of $214m consolidated. P/B on $509m of parent equity gives 20.1x and P/E 163x — both are artefacts and neither should be carried into a screen. ³ Preferred stock not tagged at a current date in XBRL; P/B computed on total stockholders' equity and therefore flatters the common. ⁴ IFRS, converted at EURUSD 1.14673 (ROIC, 2026-09-21). IFRS 17 book excludes the CSM, so AEG's and (on a different basis) LNC's P/B are not comparable to the P&C names' above them.

Read the P/E column against the cycle and it inverts. The two cheapest trailing multiples in the cluster — ALL at 6.0x and LNC at 6.8x — sit on the two earnings streams least likely to repeat: Allstate's best year in six by 2x at the exact moment the covering analyst is "not more negative on [personal lines] in my career", and Lincoln's on a series that swung from −$752m to +$3,275m to +$1,177m in three years. The most expensive, KNSL at 16.2x and 4.01x book, sits on the only earnings stream in the cluster that has compounded 5.7x on a flat share count.

Per member: variant perception, kill criteria, horizon

Bear case first, as the skill requires. The cluster's bear case is not a published short thesis — unlike the AI-capex complex, no member of this cluster carries one that this pass could find. It is the reserve cycle, and it is self-inflicted: the industry earns its way through a hard market, sets reserves conservatively, releases them into the soft market to defend reported earnings, and discovers the shortfall two to three years after pricing turned. The 2026 filings are at the start of that sequence, not the end.

Conviction: ⏳ OPEN (user).

---

§6 — What the owner should act on

1. The insurance cluster tag is doing real damage and should be split. Four of the eight are not in the P&C cycle at all. LNC and AEG are life/spread businesses whose risks are counterparty and solvency; FAF is a housing-volume business; RYAN is a fee business with an Up-C capital structure. Running S0/S4/S7 across all eight as one group — as this slot was defined — forced a single value chain onto four different machines, and the chain in §2 is honest only because it labels where each member stops being on it. Suggested: insurance-pc (ACGL, KNSL, AIG, ALL), insurance-life (LNC, AEG), title (FAF), insurance-brokers (RYAN). This is the same class of finding as the IREN/CIFR seam error in the 2026-09-20 AI-capex file.

2. FAF has an AI-capex link the watchlist does not carry. Its commercial title segment is guided to a possible record year on data-centre and energy transactions [search-summary]. That makes it a second-order beneficiary of exactly the capex the other cluster tracks — and the first name found in this repo that is long AI capex without owning a chip, a datacenter or a turbine. Worth a row in the next ai-capex cluster pass rather than an edit here.

3. The standing "no analyst-coverage count" gap now has a cheap fix. Every previous name slot recorded it as thin. Three of the eight here have a coverage count available from the same search-summary rung already used for S6 (FAF: five to seven analysts, Moderate Buy, $75–78). A name pass can close S2's coverage-count line at zero extra cost by asking for it in the S6 search.

---

Scorecard — cluster cluster-insurance

S0 Universe & fit (cluster)   ⚠️ SPLIT — KNSL ✅ 3/4 is the only pass.
                                 ACGL ⚠️2/4, AIG ⚠️2/4, FAF ⚠️2/4, AEG ⚠️2/4,
                                 RYAN ❌1/4, ALL ❌1/4, LNC ❌1/4.
                                 Owner checks (competence, exclusion list) DROPPED per README.
                                 Surplus question, falsifier and ADV written for all 8.
S4 Industry/supply map        ✅ — one chain, capital → reinsurer → carrier → wholesaler →
                                 policyholder, pricing power located on it from the issuers'
                                 own segment tables. Counterparty concentration substituted
                                 for customer concentration, with the reason stated.
S6 Scuttlebutt                ✅ 8/8 — 5 members off the board's own claim rows and the
                                 committed transcript, 3 from WebSearch (marked summary-kind).
                                 Product test / expert call ⏳ OPEN (user).
S7 Thesis + framework test    ✅ — adapted ai-capex lens table (named); cross-sectional
                                 valuation RAN (first cluster pass in this repo to close it);
                                 per-member variant perception, kill criteria and horizon
                                 written. Proxies/incentives ⬜ not examined.
                                 Conviction ⏳ OPEN (user).

➡️ Next action: names-insurance-1 (ACGL, KNSL, RYAN) carries S0/S4/S6/S7 forward from this file. Its S3 gate-minimum read should start with Arch's Q2 10-Q reserve roll-forward, because the ex-cat ex-development gap identified in §1 is the thing a name pass can either confirm or kill.

⚠️ Open: four proxies on file and unread. LNC/FAF/AEG scuttlebutt is search-summary only. No Glassdoor, Blind or channel-check evidence for any member. Title-plant economics for FAF are [background], not sourced — no XBRL tag and no filing read this pass.

Process support, not investment advice.