Research, scenarios, intake, matters — purpose-built for attorneys, CPAs, and wealth planners who structure multi-generational wealth. Primary-source citations. Currency-tracked. Refuses to fabricate.
One research engine across the full surface of an estate & tax practice —
The cited authority is always IRC, Treasury Regs, IRB-published guidance, written determinations, court cases, state codes, or treaties.
Effective dates, supersession, and amendment history are first-class fields. The pipeline filters on currency before reranking.
When the corpus does not contain sufficient authority, the system says so. It does not extrapolate from training data.
Every output ships with a versioned trail of the sources it used. Two practitioners running the same question against the same corpus get the same answer.
Title 26, Treasury Regulations, IRB-published guidance, written determinations, court opinions, state codes, treaties, and uniform acts. Refreshed continuously from the official authoritative sources.
Every substantive proposition is traced to a pinpoint authority. The system rejects anything that cannot be cited; nothing is fabricated.
Citation precision
98.4%
Hallucination rate
< 0.6%
Currency accuracy
99.1%
Effective dates, supersession, and amendment history are structural fields. Historical authority is flagged as historical — never recommended as current law.
When primary authority is insufficient, the system refuses to answer and surfaces next-step references instead. Refusal is a feature, not a failure.
Off-the-shelf retrieval pulls the whole library, then asks an AI to sift it. We do the inverse: the corpus is pre-classified, the question is routed against the classification, and synthesis only sees what is actually on-point. Higher precision, lower cost.
Every source is classified against a canonical taxonomy covering authority level, issue area, scenario, complexity, wealth band, time horizon, and jurisdiction. Free-form labels are rejected; only canonical values persist.
The question is analyzed and converted into a filter set. The corpus is narrowed deterministically before any retrieval runs — millions of passages down to thousands, in well under a second.
Synthesis sees only on-point candidates. Tighter input, higher precision, lower cost. We share the savings — routed sessions are priced below unrouted equivalents.
Three sources from the seed corpus showing how primary authority, practitioner interpretation, and superseded content are differentiated. Authority status is structural — the engine knows to cite the first two and never cite the third.
Tags: estate_tax_federal · unified_credit · portability_dsue · SLAT · BYPASS_TRUST · GST_DYNASTY · complexity:foundational · net_worth:15m_50m,50m_250m,over_250m
Tags: grantor_trust_rules · gift_tax_federal · IDGT_SALE · IDGT_GIFT · SLAT · BENEFICIARY_DEFECTIVE_TRUST · complexity:advanced
Superseded by OBBBA (Jul 2025). time_horizon=historical_superseded. Never retrieved unless query is explicitly historical.
Each strategy carries hard-coded eligibility gates, weighted signals, contraindications, a deterministic quant model, and its primary authorities. The matcher picks from this list against your client's facts — never off-the-cuff.
BENEFICIARY_DEFECTIVE_TRUST
DISCOUNTED_INTEREST_TRANSFER · C_CORP_PHC_RISK_MGMT · SECTION_83B_ELECTION · +1
CRT_UNI · CLAT_GRANTOR · DAF_CONTRIBUTION · +2
PRE_IMMIGRATION_TRUST · GILTI_HIGH_TAX_ELECTION
GRAT_ROLLING · IDGT_SALE
BYPASS_TRUST · MARITAL_QTIP
ROTH_CONVERSION_LADDER · NUA_LUMP_SUM
ANNUAL_EXCLUSION_GIFTS · 529_FRONTLOAD · SECTION_199A_OPTIMIZATION · +5
SECTION_1031_EXCHANGE · COST_SEGREGATION_STUDY · OPPORTUNITY_ZONE_DEFERRAL · +1
SLAT · ILIT · QPRT · +3
Deterministic match · instant · primary authority for every result
How the matcher works →Deterministic rules across AML, source-of-funds, OFAC, § 2036, basis consistency, FBAR/FATCA, § 877A, beneficiary drift, liquidity shortfall, stale-plan radar, and Form 706 deadlines. Every finding cites primary authority.
AML_OPAQUE
AML_LARGE · AML_FOREIGN
AML_STRUCTURED
COMPLIANCE_706
DATA_QB
ESTATE_2036
ESTATE_BENEFICIARY
ESTATE_SINGLE
ESTATE_CRUMMEY
ESTATE_IDGT
ESTATE_LIFE
ESTATE_LIQUIDITY
ESTATE_STALE
FIDUCIARY_CAPACITY
STATE_RESIDENCY
TAX_BASIS
TAX_EXPATRIATION
TAX_FBAR
TAX_REPORTABLE
Engine risk@2026-05-10· < 100ms · every finding cites primary authority
Open the full catalog →The applicable exclusion amount is $15,000,000, adjusted for inflation after 2026.
Modified by Rev. Rul. 95-58, 1995-2 C.B. 191.
The grantor's power to substitute the trustee with one related or subordinate to the grantor causes the trust to be a grantor trust under § 674.
Design partner cohort opens Q3 2026. Three firms locked, two spots remaining. Bring a representative client profile and we'll run the engine against it end-to-end — research memo, matched strategy plan, and risk findings, generated against your facts.