Change an assumption and exactly the downstream values the engine recomputed differently carry a brief gold wash. There is no hand-authored dependency map behind this: the illumination is the engine's own output diff, so it can never claim a relationship that does not exist. The parse button now narrates the steps that actually run, from reading the document through checking source evidence to running all 143 modules, with a screen-reader status line carrying the same narration.
Changelog, page 6
What shipped, and when.
A day-by-day record of what landed in Underwrite, kept by hand. Page 6 of 23.
Want to verify? The six files the decision depends on are rendered from the live source at /methodology/engine.
Show only
Filter by engine, UI, AI, ops, marketing or fixes.
2026-08-11
1 entry
The field reaches every page, and the desk starts answering. The Underwriting Field now runs site-wide with a restrained variant inside the tool and none behind the memo, and the tool gains its first microinteractions: verbatim source quotes at every parsed field, downstream values that briefly illuminate when the engine recomputes them, and a parse that narrates its real steps. The engine is untouched: Phase 3 remains at 143 modules.
2026-08-10
3 entries
The living layer ships: the site now draws the engine. A shared motion system renders the Underwriting Field behind the homepage, the SOURCE to MODEL to RISK to DECISION sequence opens the page, a short cinematic intro compresses the same story on first visit, and the new Module Universe maps all 143 modules as an interactive node field. Every animation is deterministic, budgeted per frame, and ships with a static twin under reduced motion. The engine is untouched: Phase 3 remains at 143 modules.
Every animated layer now runs on one shared frame loop with three tiers: full motion, ambient, and still. A governor watches the layer's own cost and demotes a tier whenever the rolling average spends more than 4ms a frame, offscreen effects are culled, and prefers-reduced-motion forces the still tier, where every effect renders its designed static twin instead of animating. A new automated gate drives the site in a real browser and asserts the contract from outside: 21 assertions covering input transparency, layout shift, tick cost under scripted scrolling, and byte-identical stillness under reduced motion.
The page now begins with the story the product tells: a broker sentence is detected, snaps into a model, becomes a risk environment, and resolves into the verdict, drawn on canvas and scrubbed by scroll. On first visit a cinematic intro compresses the same arc into a few seconds, always skippable, once per session, never under reduced motion. The Underwriting Field, a slow ambient drawing of the worked deal's own citations, curves, and Monte Carlo cloud, sits behind the page, and the homepage-to-tool click exits through a brief settle instead of a hard swap.
A new homepage section renders the full public module inventory as an interactive node map grouped by theme, from Macro/Rates to the Capstones. Hover names a module, selecting one opens its actual one-line description with provenance and asset coverage, and a slow gold trace walks the inventory in order, the engine running its battery. Every node is a real button, so the map is keyboard-navigable and screen-reader legible, and on small screens or under reduced motion the same inventory renders as a designed chip index instead. Nothing on the map is invented: names, counts, and descriptions come from the same inventory that /methodology/phase-3 publishes.
2026-08-10
3 entries
The document-first redesign is live across the site and the product. New type system on the Fraunces optical axis, a scroll-driven decision arc on the homepage, one owner for number style, an accessibility pass that took every measured contrast failure to zero, and zero em dashes anywhere in shipped copy. Every locked output is unchanged and the engine is untouched: Phase 3 remains at 143 modules.
The full redesign shipped: a nine-step type scale with line height bound to size, uppercase tracking that rises as size falls, four named stroke weights replacing ad-hoc opacities, three golds with stated jobs (one for edges and fills, one accessible ink for text, one deep fill for plates carrying paper text), warmed shadows, and a motion ladder with three durations and three curves. Fraunces now loads its optical-size axis, so display cuts and text cuts genuinely differ. The homepage gained a scroll-driven decision arc: the same worked deal crosses from PURSUE to PASS as the exit cap moves from 6.50% to 7.00%, because the honest version of a demo is one where the tool argues back.
Twelve local money, pct, and x formatters across eleven files were replaced by a single lib/format module. Precision is now fixed per quantity: cap rates hold two decimals (6.50%, never 6.5%), IRR holds its precision down a column, negatives take accounting parentheses at every magnitude, and one glyph, an en dash, means missing value everywhere, including the xlsx export. A lint rule now rejects any new local formatter so the convention cannot drift back.
A sweep of 9,624 rendered text nodes across five routes found 31 contrast failures and fixed all of them, including the printed memo's verdict plate and a sensitivity heat table whose light cells carried near-white text. A second harness now measures whether text physically fits the box it was given, which caught the verdict word clipping at three viewports. Separately, every em dash left in shipped copy (899 of them across 88 files) was re-punctuated by hand rules, with each replacement verified to change punctuation and nothing else. Four permanent gates guard all of it: CSS token resolution, text fit, contrast-adjacent design QA, and the formatter lint rule.
2026-08-06
1 entry
The AI assist layer moves to Claude Opus 5. The broker-memo parser, the qualitative deal review, and the IC executive summary now run on Anthropic's newest Opus model. Same forced tool use, same structured output, same privacy posture, and the deterministic engine is untouched: every number still comes from pure TypeScript with zero model involvement in the math path. Phase 3 remains at 143 modules.
Three features moved to Claude Opus 5: the broker-memo parser that reads a listing or OM and pre-fills the deal form, the qualitative deal review, and the on-demand IC executive summary. Each still runs server-side with forced tool use and a strict output schema, so the response shape and the privacy posture are unchanged. The math path did not move. The verdict, the grade, the pro forma, the debt schedule, the waterfall, and every exported figure are still produced by deterministic TypeScript, and the model only touches the extraction and qualitative layers that sit outside the engine. The model label in the AI Parser and Deal Review panels now reads claude-opus-5. No engine math changed, so the canonical Cayuga Commons numbers and the 143-module count are exactly as before.
2026-08-06
1 entry
The worked example is renamed Cayuga Commons (Illustrative Case). Same deal, same locked math, clearer label: the canonical demo now carries an illustrative identity and is not presented as a current underwriting of a specific parcel. Phase 3 remains at 143 modules.
The demo deal formerly presented as 132 Cherry Street now carries a clearly illustrative identity: Cayuga Commons (Illustrative Case), Ithaca, NY. The original name borrowed a real Ithaca address from a case study, and a real address on an illustrative demo invites the wrong inference. Every assumption and every output is unchanged: $8.5M purchase, 24 units plus 4,200 SF retail, Y1 NOI $594,066, 15.34% project IRR, PURSUE verdict, all still locked by the same regression assertions. Earlier changelog entries keep the former name as a historical record.
2026-07-13
3 entries
The underwriting journey is clearer from first click to IC memo, pricing now states the live access posture plainly, and AI parsing expands beyond PDFs to Excel pro formas, rent rolls, and underwriting workbooks. Phase 3 remains at 143 modules.
The four-step first-run tour is replaced by a compact launcher that moves analysts directly to file upload, manual inputs, or the worked-example results. The workflow rail now stays visible on larger screens, tracks the active stage, and reports extracted-field and stress-scenario status. Pricing now distinguishes what is public during evaluation from what paid plans are intended to cover, keeps the $49 Analyst, $149 Pro, and $399 Team terms intact, and states that a plan request does not create an automatic charge.
The app homepage now gives analysts three explicit starting paths: upload a source file, build from manual assumptions, or inspect a completed memo. Mobile hero typography is constrained to the viewport, supported file types are visible before entering the tool, and the welcome site uses the same calls to action. Inside the underwriting tool, a deal-aware command bar names the active deal and property type, secondary actions move into a compact menu, and a persistent five-stage workflow rail jumps directly to Source, Scenarios, Inputs, Results, and the IC memo.
The existing upload zone now accepts PDF, XLSX, XLS, XLSM, XLSB, and CSV files, with workbooks up to 32MB. Workbooks are read in the browser so large Excel binaries never cross the server request limit; underwriting summaries, assumptions, pro formas, rent rolls, debt schedules, and operating statements are prioritized before less relevant sheets. Cell values are converted to bounded, sheet-labeled source text for Claude; formulas and macros are never executed. Extracted fields retain source references for audit review, encrypted or unreadable workbooks return a clear validation error, and paste-text and PDF behavior remain unchanged. The 32MB figure in this entry described workbooks; the PDF limit it implied was never real, and the 26 August 2026 entry carries the corrected and tested figures.
2026-07-06
1 entry
Phase 3 continues: Tenant Holdover, Month-to-Month, and Post-Expiration Occupancy Risk Index lands on top of the leasing-risk stack. Where commencement sits at the front of the lease and surrender at the end, this reads the transition in between: what happens when a lease expires but the tenant remains in possession, renewal documentation is incomplete, surrender is delayed, or post-expiration occupancy disrupts re-leasing. Because the model carries no holdover clauses, month-to-month agreements, possession status, or holdover premiums, it never fabricates them: it reads their absence as a documentation and transition-control gap, books no holdover premium or possession loss, surfaces the post-expiration rent only as an observable scope, presents the near-term expiration scope and the renewal-pending proxy as distinct lenses that are never summed, and excludes the rollover and surrender exposure to avoid double-counting. 143 Phase 3 modules cumulative.
The leasing stack now covers commencement at the beginning of a lease and surrender and restoration at the end. The missing transition layer is holdover risk: what happens when a lease expires but the tenant remains in possession, renewal documentation is incomplete, surrender is delayed, or post-expiration occupancy disrupts re-leasing. A lease has an end date, but a tenant does not always leave on it: it may renew, go month-to-month, hold over past expiration, or have to be removed, and the rent the pro forma books past an expiration depends on which one happens, so when the holdover and renewal provisions, the possession status, and the transition plan are not documented, the rent past expiration is less certain than the schedule implies. The model carries no holdover clauses, month-to-month agreements, possession status, holdover notices, eviction proceedings, holdover rent premiums, executed extensions, or tenant departure confirmations, and this module never fabricates any of them: it reads missing evidence as a documentation and transition-control gap rather than confirmation that a tenant is holding over or defaulting, books no holdover-rent premium, penalty, legal cost, or delayed-possession loss, surfaces the post-expiration rent only as observable rent exposed to expiration and transition confirmation rather than a predicted loss, presents the near-term expiration scope, the renewal-pending proxy, and the possession-transition scope as distinct lenses on the rolling rent that must not be summed, excludes the rollover downtime carry and the surrender restoration cost captured by the separate modules to avoid double-counting, and preserves an unknown state for tenants without usable lease-end evidence. An underwriting screen, not legal advice, lease interpretation, eviction or possession advice, a broker opinion, accounting advice, or a property-management recommendation. Gated to commercial rent-roll deals; residential-only deals with no commercial tenant array auto-skip. Holdover posture classifies no-exposure-observed, documentation-gap, holdover-sensitive, or possession-risk from a documentation rank (abstract band complete 0 to deficiency 3, plus one on a weak amendment stack, plus one on ad-hoc or uncontrolled notice) shifted by rolling-rent volume (plus one above 60 percent rolling within the hold, minus one at or below 15 percent), where no-exposure-observed means no evidence in the modeled inputs and never confirmed absence. Occupancy-transition posture classifies planned-transition, renewal-pending, transition-watch, or unresolved from the retention band (retention-supported 0 to tenant-loss 3), plus one each on a re-tenanting-gap rollover band and a turnover-restoration surrender band. Holdover-rent posture classifies no-premium-modeled, confirmation-needed, economic-protection-watch, or unprotected-exposure, starting at no-premium-modeled because the pro forma books base rent only, raised to confirmation-needed on a weak abstract or amendment stack, to economic-protection-watch on a holdover-sensitive posture, and to unprotected-exposure on a possession-risk posture, plus one on an unsecured credit-support band. The post-expiration rent scope is the rolling rent times a holdover-scope fraction by posture (no-exposure 0.10, documentation-gap 0.25, holdover-sensitive 0.45, possession-risk 0.65); the near-term expiration scope is the WALT near-term share of rent; the renewal-pending proxy is the rolling rent times a renewal likelihood by retention posture (sticky-income 0.75, market-renewal 0.55, negotiation-sensitive 0.40, replacement-dependent 0.25); these three are distinct lenses on the rolling rent that are never added, and they exclude the rollover downtime carry and the surrender restoration cost. Six weighted drivers summing to one, scored so higher means better-documented expiration handling, a lighter near-term load, and a smoother transition: expiration and notice documentation 25% (abstract band complete 82 down to deficiency 28, plus 4 tracked to minus 8 missing critical dates, minus 4 on a weak amendment stack, plus 4 calendar-tracked to minus 6 uncontrolled notice), near-term expiration and rollover exposure 20% (WALT band term-protected 80 down to walt-compression 30, plus 4 rollover-controlled to minus 4 re-tenanting-gap, minus 2 downtime-reserve-watch), renewal and retention posture 15% (retention-supported 80 down to tenant-loss 28, plus or minus 4 on the posture), surrender and possession readiness 15% (surrender-controls-supported 80 down to turnover-restoration 28, plus or minus 4 on the surrender posture), tenant concentration and credit support 15% (concentration low 74 to high 36, plus or minus 6 on the credit-support band, plus 2 or minus 4 on credit deterioration), cash-flow cushion 10% (strong 78 to thin 30, minus 6 on a negative worst-year conversion). Bands: expiration-controls-supported (composite 72-plus with a no-exposure-observed or documentation-gap holdover posture and a planned-transition or renewal-pending transition posture), holdover-documentation-watch (58 to 71, the honest center), possession-timing-risk (46 to 57, or an unprotected-exposure holdover-rent posture or a holdover-sensitive posture under composite 55), post-expiration-occupancy-risk (under 46, or a possession-risk holdover posture under composite 58, or an unresolved transition posture under composite 55), not-assessable. Memo surface: verdict banner with band, score, and the holdover and occupancy-transition postures; four-stat headline across the possession-transition scope, its share of Y1 NOI, the near-term expiration scope, and the renewal-pending proxy; a posture strip across holdover, occupancy transition, and holdover rent; per-tenant expiration-risk rows sorted by rent share with an inferred risk tier (from rent share, escalated by expires-within-hold, a heavy holdover posture, and missing lease-end evidence), a lease-end-evidence flag that preserves an unknown state where the model lacks lease-term data, an expires-within-hold flag, and open expiration items, each framed as an item to confirm against the executed file rather than a fabricated holdover; six-driver weighted score table; operating actions tied to building an expiration and holdover register from the leases (always), starting the renewal or transition conversation ahead of each expiration (always), and underwriting the post-expiration rent from confirmed outcomes (always), escalating to confirming the holdover premium and the possession-recovery process on a holdover-sensitive or possession-risk read; diligence requests tied to the expiration dates, renewal and holdover provisions, and holdover premiums (always), the current possession status and any executed renewals or notices (always), and the surrender and re-leasing plan for each near-term expiration (always), plus a lease-file rebuild on a thin abstract; 2-4 deal implications by band with the explicit framing that on post-expiration-occupancy-risk deals heavy near-term expiration meets weak documentation and an unresolved transition and the sponsor treats the rent past expiration as unconfirmed until the renewal and holdover provisions and the possession and transition plans are in hand, prices for the transition scope, and does not book a holdover premium or a possession recovery the model cannot confirm, keeping any transition reserve separate from the rollover and surrender reserves to avoid double-counting. Module auto-skips without a commercial rent roll, and degrades gracefully when the abstract, amendment, administration, WALT, rollover, retention, surrender, tenant credit, credit support, tenant concentration, net effective rent, or cash conversion modules are absent (postures default to a documentation gap and drivers to neutral inference). Tenant holdover, month-to-month, and post-expiration occupancy risk is a deterministic underwriting screen. It is not legal advice, lease interpretation, eviction or possession advice, a broker opinion, accounting advice, or a property-management recommendation, and it never fabricates holdover status, month-to-month agreements, holdover premiums, extensions, notices, possession disputes, eviction proceedings, defaults, or departure dates.
2026-07-06
1 entry
Phase 3 continues: Tenant Surrender Condition, Restoration Obligation, and Make-Good Reserve Risk Index lands on top of the leasing-risk stack. Where commencement risk sits at the front of the lease lifecycle, this reads the opposite end: what happens when a tenant vacates, what condition the space must be returned in, whether improvements must be removed, and whether the sponsor has enough reserve for the make-good work before re-leasing. Because the model carries no surrender clauses, restoration standards, condition reports, or make-good estimates, it never fabricates them: it reads their absence as a documentation and reserve-sizing gap, books no restoration loss or tenant recovery, surfaces the restoration exposure only as a conservative scope proxy, and explicitly excludes the ongoing R&M and rollover reserves to avoid double-counting. 142 Phase 3 modules cumulative.
- Phase 3 module: Tenant Surrender Condition, Restoration Obligation, and Make-Good Reserve Risk Indexengine
The leasing stack now covers lease commencement and rent start at the front of the lease lifecycle. The missing opposite-end layer is surrender and restoration risk: what happens when a tenant vacates, what condition the space must be returned in, whether improvements must be removed, and whether the sponsor has enough reserve for the make-good work before re-leasing. A lease ends as well as begins: on surrender a tenant may owe a return of the space to a required condition, the removal of its own improvements, and the repair of alterations, or the landlord may have waived some of that, so when the surrender clause, the restoration standard, and the make-good reserve are not documented and sized, the cost of turning the space before re-leasing can land on the sponsor. The model carries no surrender clauses, restoration standards, removal obligations, landlord waivers, condition reports, make-good estimates, or tenant acknowledgments, and this module never fabricates any of them: it reads missing surrender evidence as a documentation and reserve-sizing gap rather than confirmation that restoration is owed, books no confirmed restoration loss or tenant recovery, surfaces the restoration exposure only as a conservative underwriting scope proxy requiring confirmation, distinguishes the estimated gross restoration scope from the support or recovery proxy a documented deposit or guaranty could offset and the residual sponsor funding-gap proxy, and preserves an unknown state for tenants without usable lease-end evidence. To avoid double-counting, the restoration scope excludes the ongoing repair-and-maintenance capital captured by the repair-and-maintenance shock module and the downtime carry and re-leasing capital captured by the rollover module, reading those as reserve-posture signals rather than additive dollar needs. An underwriting screen, not legal advice, lease interpretation, construction advice, cost estimation, a broker opinion, accounting advice, or a property-management recommendation. Gated to commercial rent-roll deals; residential-only deals with no commercial tenant array auto-skip. Surrender-condition posture classifies documented-control, confirmation-needed, condition-sensitive, or unresolved from the lease abstract completeness band (complete rank 0 to deficiency rank 3), escalated where the amendment stack reads unreconciled or override-sensitive (surrender terms are often modified by side letters) and where notice controls read ad-hoc or uncontrolled. Restoration-obligation posture classifies limited-scope, standard-turnover, material-restoration, or specialist-scope from the leasing-capital intensity of unrecovered tenant improvement (light 0 to heavy 3, the tenant-specific build-out most likely to require removal), escalated by a deferred-maintenance-overhang or capex-leakage R&M shock band and by a warranty-recovery or retainage-dispute closeout band. Make-good-reserve posture classifies reserve-supported, monitored-reserve, under-reserved, or sponsor-funding-gap from reserve adequacy (adequate 0 to under-reserved 2), escalated by a rollover sponsor-gap downtime posture, a thin cash conversion, and a funding-gap proxy above 12 percent of Y1 NOI. Estimated restoration exposure is the surrendering rent (the WALT rolling-within-hold share, else the rollover exposure, else a documented 0.40 default) times a conservative make-good fraction by restoration posture (limited 0.08, standard 0.15, material 0.28, specialist 0.45); the support recovery proxy is that scope times a credit-support coverage fraction (credit-backstopped 0.35, deposit-supported 0.20, thinly-secured 0.08, effectively-unsecured 0.0); the sponsor funding-gap proxy is the scope net of the recovery proxy and a conservative reserve offset (0.25 adequate, 0.10 thin, 0 under-reserved), floored at zero. Six weighted drivers summing to one, scored so higher means better-documented surrender terms, a lighter restoration burden, and a better-reserved position: surrender documentation and control 25% (abstract band complete 82 down to deficiency 28, minus 4 on a weak amendment stack, plus 4 calendar-tracked to minus 6 uncontrolled notice), lease-expiration and turnover exposure 20% (WALT band term-protected 80 down to walt-compression 30, plus or minus 4 on the rollover band, minus 3 on tenant-loss retention), restoration and capex burden 20% (leasing-capital intensity light 78 down to heavy 30, plus 4 to minus 6 on the R&M shock band, plus 2 to minus 4 on closeout, and it excludes the ongoing R&M capital to avoid double-counting), tenant support and recoverability 15% (credit-support band adequate 80 down to unsecured 28, plus or minus 4 on the posture), concentration and materiality 10% (low 78, moderate 56, high 34), reserve and cash cushion 10% (reserve adequate 76 to under-reserved 30, rollover reserve-covered plus 4 to sponsor-gap minus 6 read as a signal not re-added, cash strong plus 4 to thin minus 6). Bands: surrender-controls-supported (composite 72-plus with a documented-control or confirmation-needed surrender posture, a reserve-supported or monitored-reserve make-good posture, and no specialist scope), restoration-scope-watch (58 to 71, the honest center), make-good-reserve-gap (46 to 57, or an under-reserved posture under composite 55, or specialist scope under composite 52), turnover-restoration-risk (under 46, or an unresolved surrender posture under composite 58, or a sponsor-funding-gap posture under composite 55), not-assessable. Memo surface: verdict banner with band, score, and the surrender-condition and make-good-reserve postures; four-stat headline across the gross restoration scope, its share of Y1 NOI, the sponsor funding-gap proxy, and top-tenant exposure; a posture strip across surrender condition, restoration obligation, and make-good reserve; per-tenant surrender-risk rows sorted by rent share with an inferred risk tier (from rent share, escalated by rolls-within-hold, material restoration, missing lease-end evidence, and weak documentation), a lease-end-evidence flag that preserves an unknown state where the model lacks lease-term data, a rolls-within-hold flag, and open surrender items, each framed as an item to confirm against the executed file rather than a fabricated obligation; six-driver weighted score table; operating actions tied to building a surrender and restoration register from the leases (always), sizing a dedicated make-good reserve separate from the R&M and rollover reserves so the same dollars are not counted twice (always), and obtaining a condition report at each surrender before releasing a deposit (always), escalating to make-good estimates and tenant acknowledgment on material restoration and to funding the reserve toward the gap on an under-reserved or funding-gap posture; diligence requests tied to the surrender and restoration clauses and any waivers (always), condition reports and prior make-good estimates and tenant acknowledgments where they exist (always), and the deposit, guaranty, and credit-support detail that would backstop a restoration obligation (always), plus a lease-file rebuild on a thin abstract; 2-4 deal implications by band with the explicit framing that on turnover-restoration-risk deals heavy turnover and material restoration meet weak documentation or a thin reserve and the sponsor treats the restoration scope as unconfirmed until the surrender clauses and any tenant recovery are in hand, sizes a dedicated make-good reserve, and prices for the scope, keeping the reserve separate from the R&M and rollover reserves to avoid double-counting. Module auto-skips without a commercial rent roll, and degrades gracefully when the abstract, amendment, administration, WALT, rollover, retention, TI burnoff, credit support, R&M shock, capex closeout, tenant concentration, reserve adequacy, or cash conversion modules are absent (postures default to a documentation gap and drivers to neutral inference). Tenant surrender condition, restoration obligation, and make-good reserve risk is a deterministic underwriting screen. It is not legal advice, lease interpretation, construction advice, cost estimation, a broker opinion, accounting advice, or a property-management recommendation, and it never fabricates surrender clauses, removal obligations, restoration standards, condition reports, waivers, contractor scopes, tenant acknowledgments, or legal responsibility.
2026-07-06
1 entry
Phase 3 continues: Lease Commencement, Delivery Condition, and Rent Start Date Risk Index lands on top of the leasing-risk stack. The memo now reads commencement risk: whether tenant delivery requirements, possession conditions, rent commencement dates, outside dates, opening conditions, and free-rent burnoff could delay the contractual cash flow the pro forma books from day one. Because the model carries no commencement certificates, delivery schedules, outside dates, possession notices, or landlord-work signoffs, it never fabricates them: it reads their absence as a documentation and timing-control gap rather than a confirmed delay, books no delayed rent or default loss, surfaces the rent commencement exposure as the observable rent and concession scope requiring confirmation, and preserves an unknown state for tenants without usable date evidence. 141 Phase 3 modules cumulative.
The leasing stack now evaluates document hierarchy, amendments, concessions, lease administration, tenant retention, and capital-project execution. Those modules read the terms and the economics. The next missing layer is commencement risk: whether tenant delivery requirements, possession conditions, rent commencement dates, outside dates, opening conditions, and free-rent burnoff could delay the contractual cash flow the pro forma books from day one. A lease can be signed and still not pay: rent commences on a date that depends on the landlord delivering the space in a required condition, on the tenant taking possession, on any landlord work or buildout being complete, and on the free-rent runway burning off, so when those conditions and dates are not documented and tracked the early-period rent may start later than underwritten. The model carries no executed commencement certificates, delivery-condition schedules, outside dates, possession notices, tenant-opening confirmations, or landlord-work signoffs, and this module never fabricates any of them: it reads missing commencement evidence as a documentation and timing-control gap rather than confirmation that a delay or default exists, books no delayed rent, penalty, or default loss because no commencement data exists in the model, surfaces the rent commencement exposure as the observable rent and concession scope requiring confirmation rather than a predicted loss, and preserves an unknown state for tenants without usable date or term evidence rather than generating a date. An underwriting screen, not legal advice, lease interpretation, construction advice, accounting advice, a broker opinion, or a property-management recommendation. Gated to commercial rent-roll deals; residential-only deals with no commercial tenant array auto-skip. Commencement-date posture classifies documented-control, confirmation-needed, timing-sensitive, or unresolved from the lease abstract completeness band (complete rank 0 to deficiency rank 3), escalated where critical dates read partially-tracked or missing, where the amendment stack reads unreconciled or override-sensitive (amendments often reset a commencement date), and where notice controls read ad-hoc or uncontrolled. Delivery-condition posture classifies delivery-supported, closeout-watch, landlord-work-sensitive, or delivery-gap from the capital-project procurement band (procurement-controlled 0 to capex-delivery-gap 3) and closeout band (closeout-controlled 0 to warranty or retainage risk 2) and the scale of landlord-funded TI (elevated or heavy intensity, front-loaded or leaking recovery), defaulting to delivery-supported as the neutral read when no delivery or closeout signal exists. Rent-start posture classifies contractual-runway, concession-burnoff-watch, commencement-sensitive, or cash-flow-delay-risk from the concession intensity (low 0 to high 3) or net-effective drag fallback, compounded plus 1 each by a timing-sensitive or unresolved date posture and a landlord-work-sensitive or delivery-gap posture. Rent commencement exposure is the observable commercial rent scaled by a scope fraction: a 0.15 base, plus 0.05 (confirmation-needed), 0.15 (timing-sensitive), or 0.30 (unresolved) for the date posture and 0.05, 0.10, or 0.20 for a closeout-watch, landlord-work-sensitive, or delivery-gap posture, clamped 0.10 to 0.90; it is reported as a share of Y1 NOI and never as a confirmed loss. Concession and rent-start exposure reuses the concession module total concession exposure, else the sum of per-tenant concession term, and top-tenant commencement exposure is the largest-tenant rent share. Six weighted drivers summing to one, scored so higher means better-documented dates, readier delivery, and lower timing risk: commencement-date documentation 25% (abstract band complete 82 down to deficiency 28, plus 4 tracked, minus 3 partially-tracked, minus 8 missing critical dates, minus 4 on a weak amendment stack), delivery and closeout readiness 20% (the average of the applicable capex procurement and closeout reads, minus 4 on heavy landlord-funded TI, neutral 58 with no signal), concession and rent-start exposure 20% (concession intensity low 78 down to high 30, minus 6 on wide net-effective drag or plus 4 on tight, minus 4 on front-loaded or leaking TI recovery), lease-admin and notice controls 15% (administration-controlled 80 down to leakage 30, plus 4 calendar-tracked, minus 3 ad-hoc, minus 6 uncontrolled), tenant concentration and materiality 10% (low 78, moderate 56, high 34), stabilization and cash cushion 10% (stabilized at acquisition 78, early 66, mid-hold 50, late 38, cash strong plus 6 to thin minus 8). Bands: commencement-controls-supported (composite 72-plus with a documented-control or confirmation-needed date posture, a contractual-runway or concession-burnoff-watch rent-start posture, and no delivery gap), delivery-condition-watch (58 to 71, the honest center), rent-start-timing-risk (46 to 57, or a timing-sensitive date posture under composite 55), commencement-default-risk (under 46, or an unresolved date posture under composite 58, or a cash-flow-delay-risk or delivery-gap posture under composite 55), not-assessable. Memo surface: verdict banner with band, score, and the commencement-date and delivery-condition postures; four-stat headline across the rent commencement scope, its share of Y1 NOI, the concession rent-start scope, and top-tenant exposure; a posture strip across commencement date, delivery condition, and rent start; per-tenant commencement-risk rows sorted by rent share with an inferred risk tier (from rent share, escalated by weak date and delivery postures and by missing term evidence), a term-evidence flag that preserves an unknown state where the model lacks lease-term data, and open commencement items, each framed as an item to confirm against the executed file rather than a fabricated date; six-driver weighted score table; operating actions tied to building a commencement and critical-date register from the leases (always), requiring executed commencement certificates or estoppels that confirm the dates and delivery (always), and underwriting cash flow from the confirmed commencement dates (always), escalating to landlord-work sequencing on a weak delivery posture and explicit burnoff modeling where concession exists; diligence requests tied to the executed commencement certificates, possession notices, and delivery signoffs (always), the free-rent and concession burnoff schedule (always), and the status of any landlord work or capital project a delivery depends on (always), plus a critical-date rebuild on a thin abstract; 2-4 deal implications by band with the explicit framing that on commencement-default-risk deals unresolved dates or an incomplete delivery meet weak controls and the sponsor treats the early-period contractual rent as unconfirmed until the commencement dates, delivery signoffs, and burnoff schedules are in hand, prices for the timing scope, and makes delivery of the commencement certificates a closing condition. Module auto-skips without a commercial rent roll, and degrades gracefully when the abstract, amendment, administration, estoppel and audit, concession, TI burnoff, capex procurement, capex closeout, stabilization, net effective rent, tenant concentration, or cash conversion modules are absent (postures default to a documentation gap or a standard delivery and drivers to neutral inference). Lease commencement, delivery condition, and rent start date risk is a deterministic underwriting screen. It is not legal advice, lease interpretation, construction advice, accounting advice, a broker opinion, or a property-management recommendation, and it never fabricates commencement dates, possession dates, delivery conditions, outside dates, certificates, notices, tenant-opening status, landlord-work completion, or defaults.