Good diagnostic reasoning must also make urgency visible
Both tools identify CANOMAD. MontMedical works through the contaminated-CSF evidence in detail; OpenEvidence makes the respiratory problem more prominent.
A selected synthetic development case, published by MontMedical. The clinical fingerprints match, but the exact prompts submitted to OpenEvidence were not independently retained. This is an editorial comparison of historical answers, not a blinded efficacy trial.
Our editorial assessment
How the retained answers rate.
OpenEvidence8.2/10
MontMedical8.1/10
OpenEvidence is stronger on urgency; MontMedical is stronger on the assay boundaries.
Same five criteria and weights for both systems
Criterion and weight
OpenEvidence
MontMedical
Diagnosis and case fit 25%
9/10
9/10
Interpretation of tests and experiments 25%
8/10
9/10
Strength of the claims 20%
6/10
9/10
Decisions and next steps 15%
9/10
4/10
Clarity and teaching 15%
9/10
8/10
Why each criterion received its mark
Diagnosis and case fit
OpenEvidence · 9/10: Correctly synthesises sensory ataxia, relapsing cranial weakness and the disialosyl-IgM findings.
MontMedical · 9/10: Correct diagnosis with explicit separation from direct CNS infiltration.
Interpretation of tests and experiments
OpenEvidence · 8/10: Recognises blood contamination and the IgG-versus-IgM assay issue.
MontMedical · 9/10: Quantifies expected blood carryover and distinguishes cellular findings from soluble IgM.
Strength of the claims
OpenEvidence · 6/10: Treats Bing–Neel exclusion and some competing-test interpretations more categorically than warranted.
MontMedical · 9/10: Keeps the contamination explanation strong without claiming every alternative is excluded.
Decisions and next steps
OpenEvidence · 9/10: Makes respiratory and swallowing deterioration urgent and proposes practical escalation.
MontMedical · 4/10: Describes falling vital capacity and weak cough, but fails to make urgent escalation explicit in this retained section.
Clarity and teaching
OpenEvidence · 9/10: Strong opening and clear account of the clinical pattern.
MontMedical · 8/10: Clear assay reasoning; clinical priorities are buried among additional findings.
AI-assisted editorial judgment of these historical texts, published by MontMedical. Read the rubric, anchors and limits. These are answer-quality scores, not measured diagnostic accuracy or patient benefit.
The actual retained wording
OpenEvidence vs MontMedical.
These are verbatim passages from the retained answers, not rewritten examples. Bold, link formatting and line breaks have been removed. Each labelled passage is contiguous; an excerpt may begin within a sentence. Our interpretation and scores are separate.
OpenEvidence
Retained 2026-10-01
Opening priority
CANOMAD (chronic ataxic neuropathy, ophthalmoplegia, IgM paraprotein, cold agglutinins, and disialosyl antibodies) best explains this presentation, and the patient is now in an acute relapse with impending respiratory failure requiring urgent escalation.
Exclusion claim
Bing–Neel syndrome — effectively excluded by the contamination pattern on repeat atraumatic LPs and absence of intrathecal IgM synthesis, as above.
MontMedical
Retained 2026-10-01
Respiratory finding
Current findings include coughing with water, inability to sustain the head upright and a weak cough. Forced vital capacity has fallen from 76% to 48% predicted, while oxygen saturation and arterial carbon dioxide remain normal.
Contamination calculation
Calculated proportional admixture predicts approximately 8.3 circulating clonal cells in the tested aliquot: (4,000/4,800,000) × 5 × 2,000, compared with nine reported events. Matching phenotype, mutation and clonotype therefore need not represent independent evidence of CNS residence; all can accompany the same circulating clone.
The case and the decision
A synthetic patient has sensory ataxia, relapsing eye and bulbar weakness, monoclonal IgM and independently confirmed disialosyl IgM. A traumatic CSF specimen contains a few matching clonal cells, while subsequent adequate atraumatic samples are negative. Swallowing, cough and vital capacity are deteriorating.
What MontMedical explains particularly well
Its proportional-admixture calculation predicts about 8.3 circulating clonal cells in the tested CSF aliquot, compared with nine detected events. That makes blood carryover quantitatively sufficient to explain the sparse cells. Matching phenotype, mutation and clonotype may all arrive in the same contaminating blood; they are not automatically three independent demonstrations of CNS residence.
This interpretation is consistent with Bing–Neel guidance warning about blood contamination in CSF cellular and molecular testing. It does not turn negative follow-up samples into a universal exclusion test. [1]
Where OpenEvidence is stronger
OpenEvidence opens with the acute relapse and impending respiratory failure. The retained MontMedical version places the deteriorating cough and vital capacity under additional findings without equally prominent immediate-action wording. On presentation of urgency, OpenEvidence is stronger in this pair.
That matters: an accurate disease name and a careful assay explanation are not enough if the action-relevant finding is easy to miss.
The shared diagnosis and the separate causal question
The clinical pattern is well aligned with the published CANOMAD phenotype. Both tools recognise that an earlier IgG-only anti-GQ1b assay does not answer the later IgM question. [2]
MontMedical also separates the established marrow clone from proof that its reconstructed antibody binds the relevant gangliosides. Testing the clone product would address that narrower causal link. Our preferred design combines that precision with OpenEvidence's clearer prominence of urgency.
Retained answer dates: MontMedical 2026-10-01; OpenEvidence 2026-10-01. Excerpt wording was checked against the retained answers; formatting was removed. These versions were selected during development. Complete original run settings and identical submitted prompts are not documented.
This article examines the MontMedical Pro beta using retained diagnostic answers.