How Luhn works
A public intelligence record for the multifamily housing and proptech sector. We compress fragmented public evidence into named narratives, track our predictions accountably over time, and publish the discipline as well as the conclusions.
Every Luhn brief on the public record carries these nine elements. None are optional. The structure is the trust system.
- 01Narrative claimThe compressed thesis. One sentence implying consequence.
- 02Evidence boundaryWhat the brief does not claim. The line between observed events and posture, named explicitly.
- 03Source trailEvery claim traces to an inspectable source row. Publisher, date, what it supports.
- 04Counter-signalA substantive disconfirming condition. What would weaken or invalidate the thesis.
- 05Maturity bandConfidence as a lifecycle position — Early Signal, Strengthening, Confirmed Trend, High Conviction.
- 06Narrative statusA separate axis from maturity — Emerging, Strengthening, Confirmed, Weakening, Invalidated.
- 07Watch nextThree to five monitorable events. Outcomes published — including the ones that fail to materialize.
- 08Operator implicationWhat the narrative means for an operator with direct exposure. Scoped, not totalizing.
- 09Revision historyEvery editor-approved change to a published brief. Always rendered, even when empty.
Every brief states what it does not claim. The line between observed or executed events and posture or risk is named explicitly, not implied.
“No criminal indictment has been filed. This brief evaluates enforcement posture, not completed prosecution.”
Every brief carries at least one substantive disconfirming condition — what would weaken or invalidate the thesis. Never a throwaway hedge. The counter-signal is the firewall against over-narrativizing.
Confidence is a lifecycle position, not a number. Early Signal → Strengthening → Confirmed Trend → High Conviction. Each band has to be earned by evidence. The raw scoring composition — consequence, source quality, novelty, contradiction, velocity, cross-source confirmation — is visible on every brief page. Transparent, not a black box.
A separate axis from maturity. Where a thesis sits in its lifecycle: Emerging → Strengthening → Confirmed → Weakening → Invalidated. Briefs move through these states in public as evidence arrives or fails to.
Every brief carries three to five specific, monitorable events. Both confirming and disconfirming. Outcomes are published — including the ones that don't materialize. Non-events are intelligence too.
Luhn ingests from primary regulator filings, major wires, sector trade publications, and selected legal and research aggregators. Source quality is weighted by publisher proximity to the source event, not by feed configuration.
- Primary — regulator filings, court records, direct corporate disclosures. Highest weight.
- Wire — major institutional news wires with proximity to the source event.
- Trade — sector-specific publications with domain depth.
- Secondary — aggregators and downstream commentary. Weighted lower, used only as confirmation.
The Editorial Judge — the final filter between AI proposals and published briefs — is itself on the record. Every recommendation it makes is tracked against the editor's eventual decision. Agreement rate, per-principle citation frequency, and override patterns are published as the calibration record accumulates.
Luhn judges its own judges, publicly. The track record is the credibility.
Every editor-approved change to a published brief is recorded as a revision on that brief's page. Briefs that have not been revised display "No revisions published." — the empty revision history is itself a public statement.
Corrections do not silently overwrite. If a published claim turns out to be wrong, the original text and the correction both remain visible in the brief's revision history.
We don't auto-publish anything reader-facing. AI drafts; an editor reads, calibrates, and publishes. The discipline is the product.
We don't synthesize aggregate figures unless one source states them. We don't imply events have happened when only authorization or referral exists. We don't use reckoning, collapse, crisis, fracture, criminal, exposure, wave, trap, or reset unless the evidence specifically earns them.
The standards encoded
The standards above are not abstract. They are encoded as a published constitution every brief must pass before it reaches the public record. Updates to these standards govern every draft, review, and published brief.
No Hedging Verbs
CriticalBanned verbs that smuggle uncertainty into prose without earning it: suggests, indicates, appears, seems, may signal, could imply, points to, hints, reflects (as a hedge). State observable behavior and the direct implication. Let the evidence carry uncertainty, never the prose. If the evidence is uncertain, name the uncertainty directly — do not paper over it with a soft verb.
Banned:“suggests”, “indicates”, “appears”, “seems”, “may signal”, “could imply”, “points to”, “hints”, “reflects (as hedge)”
- “Lenders are protecting prime assets and letting weaker portfolios reprice.”
- “Three operators raised marketing spend in the same quarter.”
- “The divergence suggests price discovery is forcing hands.”
- “The data may signal a broader operator-margin trap.”
Compression Over Hedging
CriticalSharp, short, declarative. The house voice compresses complexity into a phrase that implies consequence. Hedging language and consultant-ese are always weaker than the plain observable fact. Cut every word that does not add evidence or implication.
Banned:“it is worth noting”, “in some ways”, “to some extent”, “broadly speaking”, “arguably”
- “Free rent returns.”
- “The operator margin trap.”
- “It is worth noting that, in some ways, operators may be facing a broadly difficult margin environment.”
No Clickbait, No Fake Certainty
CriticalHeadlines and titles imply consequence without inflating it. Never use clickbait constructions ("you won't believe", "this changes everything", "the one thing"). Never assert certainty the evidence does not earn. Sharpness without false confidence.
Banned:“you won't believe”, “this changes everything”, “the one thing”, “shocking”, “unprecedented (when not literally true)”
- “THE RENT WALL”
- “LENDERS STOP PRICING REALITY”
- “You won't believe what just happened in multifamily.”
- “The shocking truth about RealPage.”
No Meta-Reference
CriticalDo not write about the sources or the reporting itself. Prose stays inside the narrative. "Evidence stories reference cap rate compression…" or "the reporting suggests…" pulls the reader out of the story and signals weak conviction. Cite what is happening directly.
Banned:“evidence stories reference”, “the reporting suggests”, “according to the signals above”, “the data we gathered shows”
- “Cap rates compressed 80 bps across reporting operators in Q3.”
- “Three operators raised marketing spend in the same quarter.”
- “Evidence stories reference cap rate compression.”
- “The reporting suggests operators are under pressure.”
Earned Force
Criticalreckoning, collapse, crisis, fracture, criminal, exposure, wave, trap, reset are overused industry words. Use only when the evidence specifically earns that strength — and when it does, do not dilute. Earned force, not less force. The default plainer word is almost always correct.
Banned:“reckoning (default)”, “collapse (default)”, “crisis (default)”, “fracture (default)”, “wave (default)”, “trap (default)”, “reset (default)”
- “Three operators cut marketing spend by half.”
- “The DOJ filed criminal charges Tuesday.”
- “A reckoning for the multifamily sector.”
- “The proptech collapse.”
- “The operator margin trap (when evidence is one quarter of soft data).”
No Default Catastrophe
CriticalDo not force a bearish frame. If the evidence points to expansion, adaptation, recovery, operator innovation, or underestimated strength, label it that way. Tone must match evidence direction, not a default doom reflex. Upside is a valid narrative — hidden strength, operator adaptation, emerging winners, margin recovery, demand recovery, capital reopening — and the judge weights upside-evidence stories with the same rigor as downside ones.
- “Three large operators absorbed the rent compression without margin loss.”
- “Vendor consolidation is creating advantage for the survivors.”
- “The market is collapsing (when evidence shows mixed conditions).”
- “Operators face a reckoning (when evidence shows adaptation).”
No Aggregate Fabrication
CriticalNever assert a summed or aggregate figure — dollars, counts, percentages — unless one source states that exact number. Do not total separate figures from multiple stories and present the sum as fact. If multiple figures appear in evidence, list them or describe them as a combined approximation, never as a single asserted total.
Banned:“$XB total (when no single source states $XB)”, “across X operators (when summing separate single-operator figures)”
- “Three publicly traded operators reported individual writedowns of $90M, $110M, and $68M.”
- “The combined figure across reporting operators is approximately $268M.”
- “Operators have written down $268M.”
- “The sector has lost $1.2B in market value.”
Confirmed vs Pending Evidence
CriticalNo inevitability language unless confirmed. Risk is not certainty. The label and prose must not imply an event has happened unless the evidence proves it. "Operators face refinancing risk" is allowed; "operators have defaulted" requires actual defaults in evidence.
Banned:“will collapse”, “inevitable”, “is about to”, “has begun (without confirming evidence)”
- “Operators with 2025 maturities face refinancing risk on portfolios marked above market.”
- “If the Q4 trend holds, three additional operators cross into covenant breach territory.”
- “Operators have defaulted (when only delinquency rose).”
- “The collapse has begun (when only one indicator weakened).”
No Totalizing Language
CriticalBanned unless evidence supports a literal universal claim: every operator, no peer can match, all operators, the market has reset, this changes everything, now competes against [X], no one can do [X]. Replace with scoped phrasing: few operators can match at current scale, raises the competitive bar for similarly exposed operators, changes the comparison set for [scoped group].
Banned:“every operator”, “no peer can match”, “all operators”, “the market has reset”, “this changes everything”, “no one can do”
- “Few operators can match at current scale.”
- “Raises the competitive bar for similarly exposed REITs.”
- “Changes the comparison set for vertically integrated operators.”
- “Every operator now faces this.”
- “The market has reset.”
- “This changes everything.”
Operator Scope Required
CriticalName the defined group the evidence implicates. Not a broad ambient claim. "Operators with direct exposure to RealPage-style pricing workflows" is precise. "The industry" is not. Scope is part of credibility.
Banned:“the industry”, “the sector (when broader implication is unearned)”, “everyone in multifamily”
- “Operators with direct exposure to RealPage-style pricing workflows.”
- “Class B garden-style portfolios in Sun Belt MSAs.”
- “Refinancing-exposed operators with 2025 maturities.”
- “The industry is facing this.”
- “The whole sector is exposed.”
Signaled vs Acted
CriticalDistinguish proposed, signaled, referred, authorized, or filed from acted, ruled, charged, completed, or executed. A motion is not a ruling. A referral is not an indictment. An authorization is not a transaction. The judge weights this strictly — premature outcome attribution is a critical violation.
Banned:“was charged (when only referred)”, “ruled (when only motioned)”, “completed (when only authorized)”
- “The DOJ filed a pre-trial dismissal motion; the court has not ruled.”
- “The board authorized up to $500M; no transactions have been disclosed.”
- “The court ruled in favor of dismissal (when only the motion was filed).”
- “The company spent $500M (when only authorized).”
Richard Millar Devens used the phrase business intelligence in print in 1865, describing a banker named Sir Henry Furnese who gathered information before his competitors and acted on it. Hans Peter Luhn formalized the practice in his 1958 IBM paper. We sit in both traditions: the practice as old as commerce, the discipline as old as computing, the accountability as new as we are willing to demonstrate.