Series 1 · Trusts and estates

The Second Codicil

Q19Four modelsSix blind pairingsMaterial-outcome scoring

Why this result mattered

Functional tie

The principal comparison placed Qwen 3.8 and Claude under concealed A/B labels.

After ordinary verification, both readers preserve the evidence, pursue the brokerage-account and exploitation claims without triggering the no-contest clause, seek mediation, and evaluate a timely codicil contest based principally on undue influence. The record supplies neither litigation deadlines and costs nor reliable success probabilities establishing that A’s staged filing or B’s earlier codicil filing produces a materially better outcome.

Benchmark materials

Prompt and sealed key

The key is an audit reference. It was not shown to the contestant systems and did not determine the comparative verdict.

Read the original matter prompt

Question 19: The Second Codicil

Your role

You represent Claire Andersen, daughter of the late Henrik Andersen, in evaluating whether to challenge her father's estate plan. She wants a clear-eyed assessment of her chances, the money at stake, and the risk that contesting costs her what she already has.

Governing law

  1. Testamentary capacity requires that, at the time of execution, the testator knew the natural objects of his bounty, the nature and extent of his property in a general way, and the disposition he was making. The standard is lower than capacity to contract. A diagnosis of dementia does not itself defeat capacity; execution during a lucid interval suffices.

  2. Undue influence overcomes the testator's free will. A presumption of undue influence arises when a person in a confidential relationship with the testator actively participated in procuring the instrument and receives a substantial benefit under it. The presumption shifts the burden to the beneficiary to show the disposition was the testator's free act.

  3. Active procurement includes recommending the attorney, arranging the execution, instructing the attorney, supplying information about assets, or isolating the testator from family.

  4. A no-contest clause is enforceable, but forfeiture is not imposed against a beneficiary who contests in good faith and with probable cause.

  5. Assets with payable-on-death designations or survivorship features pass outside the will. A joint or payable-on-death account opened for the depositor's convenience, rather than with donative intent, may be recovered by the estate on clear and convincing evidence.

  6. Claims of financial exploitation of a vulnerable adult may be brought against a caregiver independent of the probate contest, with remedies including recovery of transferred property and fee shifting.

  7. A challenge to a nonprobate transfer is not a "contest" of the will unless the instrument's no-contest clause expressly reaches nonprobate transfers.

Facts

Henrik died at 88, leaving an estate of approximately $2.9 million: a house appraised at $850,000, brokerage and bank assets, and personal property. His 2019 will divided everything equally between his two children, Claire and Erik.

Three years before his death, Henrik hired a live-in caregiver, Sofia Marsh. Fourteen months before his death, Henrik was diagnosed with vascular dementia. His physician's notes across the following year describe "fluctuating orientation," "good days and bad days," and one note five weeks before the codicil: "clear and conversant today, recalled family details."

Ten months before his death, Henrik executed a codicil leaving the house to Sofia. The appointment was scheduled by Sofia with a lawyer Sofia located, whom Henrik had never used. Sofia drove Henrik to the office and, per the drafting lawyer's memo, "waited in the reception area during the entire conference." The memo continues: "Client alert, oriented, adamant about rewarding Sofia's care. Reviewed assets from a list client brought." The asset list is in Sofia's handwriting. The codicil added a no-contest clause providing that any beneficiary who "contests this will or codicil" forfeits all benefits under them. The clause does not mention nonprobate transfers.

Eight months before his death, Henrik's $700,000 brokerage account was retitled payable on death to Sofia. The bank's new-accounts memo records Henrik saying the change was "so Sofia can handle my bills if I'm laid up." Sofia was present at the branch.

Claire had been estranged from Henrik for roughly two years following a dispute, though her phone shows unanswered texts attempting reconciliation during his final year. Erik, who visited monthly, says Henrik told him last spring, "Sofia takes care of everything now, she even tells me who calls."

Under the current instruments, Sofia takes the house ($850,000) and the brokerage account ($700,000); Claire and Erik split the residue, roughly $675,000 each. If the codicil and the account transfer both fail, Claire and Erik split approximately $1.45 million each.

Assignment

Analyze:

  • Capacity: how the dementia diagnosis, the fluctuating notes, the "clear and conversant" entry, and the drafting lawyer's memo interact under the lucid-interval rule.
  • Undue influence: walk each element of the presumption through these facts, including every act of active procurement, and assess whether Sofia can rebut.
  • The drafting lawyer's memo: which parts help Sofia, which help Claire, and how the handwritten asset list cuts.
  • The brokerage account: the convenience-account theory, the burden of proof, Sofia's presence at the branch, and the bank memo.
  • The no-contest clause: exactly what Claire risks by contesting the codicil, how the probable-cause exception applies to this record, and whether attacking the account transfer triggers the clause at all under its text.
  • The independent exploitation claim: elements, remedies, fee shifting, and how it interacts with the probate strategy.
  • The expected-value arithmetic: Claire's position if she does nothing, if she contests and wins on both, wins on one, or loses with and without probable cause. Show the numbers.
  • Settlement architecture: what a mediated resolution with Sofia and Erik should look like, and what leverage the exploitation claim adds.
  • The recommended course, stated plainly, with the first three concrete steps.


Read the sealed key

Q19 key (Second Codicil)

  1. Capacity is the weaker attack: the lucid-interval rule, the "clear and conversant" note five weeks prior, and the drafting lawyer's contemporaneous memo make a capacity-only contest likely to fail. Missed: the case is staked on the losing theory.
  2. Undue influence is the winning frame: confidential relationship (live-in caregiver), substantial benefit (the house), and active procurement on at least four supplied indicia (found the lawyer, scheduled, drove, supplied the asset list in her own handwriting), raising the presumption and shifting the burden to Sofia. The isolation evidence ("she even tells me who calls") compounds it. Missed: the burden-shift, the case's engine, goes unclaimed.
  3. The lawyer's memo cuts both ways and the answer must say so: "alert, adamant, waited outside" helps Sofia on capacity and execution; the caregiver-arranged appointment and her handwritten asset list feed procurement. Missed: one-sided reading of the central document.
  4. The brokerage account is a separate, cleaner claim: the bank's own memo ("so Sofia can handle my bills") is direct convenience-account evidence rebutting donative intent, subject to the clear-and-convincing burden. Missed: $700,000 analyzed under the wrong instrument or forgotten.
  5. The no-contest clause: by its text it reaches contests of "this will or codicil," not nonprobate transfers, so the account challenge is safe; the codicil contest risks Claire's roughly $675,000 residuary share only if brought without probable cause, and this record (presumption triggered on documented facts) is close to a definitional probable-cause case. Missed: the client is either scared off a protected claim or walks unwarned into forfeiture analysis.
  6. The exploitation claim adds recovery of transferred property and fee shifting against Sofia independent of probate, and settlement pressure. Missed: leverage left unused.
  7. The arithmetic, stated: do nothing, about $675,000; win both, about $1.45 million; win the account only, about $1.025 million; lose with probable cause, still about $675,000; lose without probable cause on the codicil contest, near zero from the estate. The asymmetry favors contesting given the probable-cause strength. Missed: the client decides without the numbers.
  8. Recommended course: challenge the account transfer first or simultaneously (outside the clause), bring the codicil contest on the undue-influence presumption, file or threaten the exploitation claim, and drive to mediation with Erik aligned. First steps: preserve the bank and drafting-lawyer files, retain a geriatric-capacity expert, demand the caregiver's records. Missed: sequencing that triggers risk before leverage exists.
Blind evaluation record

Direct judge report

The report below is preserved from the Qwen 3.8 versus Claude verdict record.

Read the preserved judge report
TIE, FUNCTIONALLY EQUIVALENT.

Basis. After ordinary verification, both readers preserve the evidence, pursue the brokerage-account and exploitation claims without triggering the no-contest clause, seek mediation, and evaluate a timely codicil contest based principally on undue influence. The record supplies neither litigation deadlines and costs nor reliable success probabilities establishing that A’s staged filing or B’s earlier codicil filing produces a materially better outcome.

[A] rejected: numerical success probabilities, litigation costs, and resulting expected values — record lacks empirical probability and cost inputs.
[B] rejected: numerical success probabilities, litigation costs, and resulting expected values — record lacks empirical probability and cost inputs.
[B] rejected: chronic fluctuating dementia shifts the lucid-interval burden to Sofia — supplied governing law contains no such burden-shifting rule.
[B] rejected: Sofia “hits all five” procurement markers, including instructing the attorney — record contains no evidence that she instructed the attorney.
[B] rejected: the bank memo is “close to dispositive” and the account claim has a 65–75% success probability — record supplies a clear-and-convincing burden but no basis for that quantified prediction.
[B] rejected: Erik-only filing, court-approved settlement, subpoenas, or lis pendens necessarily eliminate or preserve specified risks — record lacks the procedural and indirect-contest rules required for those conclusions.
Failed candidate: immediate versus reserved codicil contest — concrete consequence and record support failed because no filing deadline, cost, or outcome probability is supplied.
Failed candidate: differing assessments of claim strength — record support failed for the quantified probabilities, while both answers identify the same operative evidence and burdens.
Failed candidate: differing settlement floors and target ranges — record support failed because caregiving value, litigation costs, and Sofia’s settlement preferences are absent.
Failed candidate: B’s additional procedural measures — likelihood failed because the record does not establish their availability, necessity, or effect.

Hard stop. The required sentence cannot be completed with a concrete, record-supported consequence.
Complete comparison record

All six pairings

Every pair of the four model answers was judged independently under the same evaluation instrument.

Open the six head-to-head results
PairingResultWhy it matteredRecord
Qwen 3.8 27B vs Claude Opus 5Functional tieAfter ordinary verification, both readers preserve the evidence, pursue the brokerage-account and exploitation claims without triggering the no-contest clause, seek mediation, and evaluate a timely codicil contest based principally on undue influence.Judge record
Qwen 3.8 27B vs DeepSeek V4-Flash-0731Qwen 3.8 winAnswer A recommends pursuing the nonprobate account and exploitation claims first while reserving the codicil contest; Answer B recommends a combined action if mediation fails.Judge record
Qwen 3.8 27B vs Qwen 3.5 27BQwen 3.8 winAnswer A recommends immediately contesting the codicil while assuring Claire that proceeding correctly effectively protects her existing $675,000; Answer B instead first pursues the independently actionable brokerage transfer, which cannot trigger the clause as written.Judge record
DeepSeek V4-Flash-0731 vs Claude Opus 5Functional tieBoth answers place Claire on the same material path: preserve evidence, challenge the POD transfer without forfeiture exposure, document probable cause before contesting the codicil, pursue exploitation remedies, and seek a global mediated settlement.Judge record
DeepSeek V4-Flash-0731 vs Qwen 3.5 27BFunctional tieBoth answers place Claire on the same material path: preserve and obtain the evidence, pursue the strong and no-contest-safe POD challenge, assert the codicil challenge with documented probable cause, use the exploitation claim as settlement leverage, and seek a global resolution before or alongside…Judge record
Qwen 3.5 27B vs Claude Opus 5Functional tieBoth answers direct Claire to preserve the evidence, challenge the codicil primarily for undue influence, pursue the POD and exploitation claims, document probable cause, coordinate with Erik, and seek a global settlement.Judge record
Rejected-point diagnostic

Judge-attributed rejected points for this matter across all six pairings:

Qwen 3.8 27B3
DeepSeek V4-Flash-07318
Claude Opus 512
Qwen 3.5 27B9

This is a review-burden diagnostic, not a universal hallucination measure.

Available reasoning traces