Homeo AI vs Aquanta simili.ai: Bayesian Beta or Review Copilot?

Quick answer

Choose Aquanta simili.ai when the narrow test is a free multilingual beta that accepts symptom text and returns Bayesian-ranked remedy suggestions with confidence information for practitioner review. Choose Homeo AI when the main job is to preserve a completed case narrative, inspect proposed symptom and rubric structure, open supporting source passages, correct the differential, and keep the accepted reasoning available for follow-up.

Aquanta links its public product site to a live simili.ai beta. On 24 August 2026, the beta described free access during testing, a maximum of ten cases per day, support for desktop browsers, and limited mobile use. Beta status matters. Availability, limits, output format, hosting, account rules, and continuity can change without the contractual stability expected from a mature clinic system.

Aquanta also publishes a 2026 research preprint. That document is useful because it exposes a technical rationale and limitations that are usually absent from a product page. It does not independently validate the live simili.ai beta, prove therapeutic efficacy, or establish that a ranked remedy is clinically correct. A buyer should keep the product claim, the vendor-authored study, and independent clinical validation as three separate evidence categories.

What the current beta offers

Aquanta describes a Bayesian statistical engine combined with AI-assisted review. The site says its knowledge base includes hundreds of thousands of symptom-remedy relationships and that the service supports multiple languages. The output is presented as ranked guidance rather than a diagnosis. These are vendor statements to test. They do not establish the completeness of the underlying repertory, the source and edition of every relationship, or the calibration of a confidence value.

Bayesian ranking can be useful when its inputs, assumptions, and likelihood updates are visible enough to challenge. Ask what a displayed confidence measure means. It may represent a normalized ranking, model certainty, rubric coverage, or another internal quantity. It should not be read as the probability that a remedy will help a patient. Repeat a case after small wording changes and record whether ranking changes are stable and explainable.

The current daily limit makes a disciplined trial possible. Ten cases per day is enough for a small de-identified test set, but it may be insufficient for a busy clinic or repeated correction. Confirm whether editing and rerunning a case consumes another allowance, whether accounts retain cases, what export exists, and what happens when the beta ends. Free access is a temporary commercial condition, not a long-term price.

What the preprint does and does not show

Aquanta's vendor preprint compares historical homeopathic proving descriptions with modern toxicology signs using language-model embeddings. It describes a larger remedy set, a gated analysis subset, statistical correction, and a group of remedies that passed its significance threshold. The research question is about semantic coherence between two bodies of text.

That is not the same question as whether simili.ai selects a clinically appropriate remedy from a real case. The paper does not test patient outcomes, therapeutic efficacy, the live interface, error rates in narrative interpretation, practitioner correction, follow-up decisions, or comparative safety. It should not be cited as evidence that the product works clinically.

The paper's own limitations are important for procurement. Semantic similarity is not clinical validity. The selection gate changes which remedies enter the final analysis. A language model was involved in the research pipeline, which creates a possible circularity when related language representations are compared. The embedding model was not medically validated for this use. Code and data are described as available on request rather than as an independently reproducible public package in the reviewed materials.

This does not make the research worthless. It makes the correct claim narrower: Aquanta provides a vendor-authored statistical and semantic rationale that can inform questions for a product trial. It does not remove the need to inspect sources, errors, uncertainty, and clinical responsibility.

How Homeo AI differs

Homeo AI centers the review path rather than a single Bayesian rank. The practitioner starts with an approved narrative, checks proposed symptom structure and rubric mappings, edits or rejects them, opens linked passages, reviews remedy differences, and preserves the accepted rationale for follow-up. Its value depends on whether the transformation can be reconstructed and corrected.

Homeo AI should not be described as using Aquanta's Bayesian method or validated by Aquanta's preprint. Aquanta should not be assumed to reproduce Homeo AI's exact correction trail merely because it accepts symptom language, uses AI, or presents confidence information. The two products make different promises that require different tests.

If the clinic wants a fast ranked list from a constrained beta, Aquanta may be the clearer experiment. If the clinic wants another practitioner to see why ordinary wording became a symptom and rubric structure, challenge the source passages, and compare follow-up changes, Homeo AI may fit the narrower review job.

Data and governance questions for a beta

Aquanta's public footer describes the product as educational and not diagnostic. The reviewed root and about pages did not expose a separate public privacy or terms link for the beta. Do not infer that no internal controls exist. The absence of a reviewable public document means a practitioner needs a written answer before entering any information that could identify a person.

Ask which entity operates simili.rmlx.cloud, where prompts and accounts are hosted, which subprocessors and model providers receive content, whether inputs or outputs are retained or used for training, how long logs and backups remain, and how export and deletion work. Ask whether the product stores cases by default, whether a user can remove them, what staff can inspect for support, and what notice will be given before the beta changes or closes.

Use only completed, de-identified cases. Remove names, dates of birth, contact details, record numbers, exact visit dates, photographs, addresses, uncommon locations, rare occupations, and combinations that could identify a person. Keep any re-identification key outside the test. Limited mobile support also means the clinic should verify screen privacy, local browser storage, session expiry, and device loss on every permitted device.

A reproducible practitioner test

Prepare eight completed, de-identified cases already reviewed by the same practitioner. Include two straightforward cases, two close differentials, one sparse case, one case with ambiguous wording, one follow-up, and one case that should produce explicit uncertainty. Keep an answer packet outside both tools containing the original wording, accepted symptom structure, sources, rejected interpretations, final differential, and follow-up changes.

In simili.ai, record the input language, interpreted symptoms, ranked remedies, confidence output, visible explanation, source identity, time, and daily allowance consumed. Repeat two cases with harmless wording changes and one with a material correction. Check whether the changes in rank are understandable, whether the product distinguishes missing evidence from low confidence, and whether another practitioner can reconstruct the result without relying on the vendor's preprint.

In Homeo AI, enter the same narratives. Record every proposed symptom and rubric accepted, edited, or rejected, each passage opened, how the remedy differential changes after correction, and whether a second practitioner can reproduce the accepted review. Reopen the follow-up and verify that the earlier rationale remains visible.

Score both products on narrative fidelity, source traceability, correction time, uncertainty handling, repeatability, export completeness, data deletion, device behavior, and continuity after the beta. Do not score agreement with a historical prescription as proof of accuracy. A ranked match can result from incomplete information or an unsupported mapping.

Known limits and clinical responsibility

A Bayesian label does not make uncertain inputs reliable. A confidence value is not clinical certainty. A large relationship count does not prove source quality, coverage, or the right interpretation for a particular case. A vendor preprint is not independent product validation, and semantic coherence is not therapeutic efficacy.

Homeo AI can also misunderstand a narrative, suggest an unsuitable rubric, omit relevant evidence, or create a plausible differential that needs correction. Source links and edit controls improve reviewability; they do not guarantee a correct conclusion.

Neither product replaces diagnosis, referral, an emergency-care decision, practitioner judgment, prescription, potency, dosage, or follow-up responsibility. Urgent, severe, or rapidly worsening symptoms require appropriate medical assessment. The practitioner remains accountable for the original case, every accepted interpretation and source, and every clinical action.

Conditional verdict

Choose Aquanta simili.ai if the practitioner specifically wants to test Bayesian remedy ranking in a free multilingual beta and receives satisfactory written answers about data use, sources, confidence meaning, export, deletion, and post-beta continuity. Treat the research paper as a source of test questions, not as proof of live product accuracy.

Choose Homeo AI if the priority is a reconstructable path from narrative through corrected symptom and rubric structure to inspected passages, remedy differentiation, and follow-up. If both are tested, keep the same answer packet and compare measured correction burden, traceability, uncertainty, and governance rather than the apparent authority of a ranked number.

Official sources checked

• Aquanta official product page
• Aquanta official about and methodology context
• Aquanta official live simili.ai beta
• Aquanta Labs 2026 vendor preprint

Product facts reviewed 2026-08-24. Confirm current features, prices, source libraries, and terms with the vendor.