Frame the question
Restates what is being asked and what a good answer must settle, so you can confirm it understood the question you meant.
You can challenge: the framing itself.
6 Shishyan Thinking
Not just an answer. A mini report you can audit.
Our deep-tech research wing built a thinking process inspired by how doctors and medical researchers reason. For every task or question you get how it thought, why a decision was suggested, the evidence, the probabilities and the context.
- Transparent reasoning
- Evidence you can check
- Probabilities, not certainty
- Decision support; you decide
Onboarding happens through this form. There is no separate sign-up or login.
What you get
A mini report for every question
Open each part of the report below. Every claim points back to a record or a source, and every step can be challenged.
Question (synthetic): HbA1c has risen to 8.1 % on metformin 500 mg twice daily. What are the reasonable next steps?
Suggested direction
Optimise the current agent first: confirm adherence and dose, review diet and activity, and recheck in the usual interval. If the target is still missed, discuss intensifying therapy.
Why
The rise is modest, the dose has headroom, and adherence has not yet been confirmed, which is the commonest cause of an unexplained rise.
Likelihood of each option being appropriate
Probabilities express the relative likelihood of each option given this record and the guidance, not a guarantee. Confidence: moderate, because adherence is unconfirmed.
Evidence from the record
- LabHbA1c 8.1 %, lab report photo, verifiedrecord #R-204
- MedicineMetformin 500 mg twice daily, from strip photorecord #R-198
- GapNo adherence check on filemissing
Evidence from guidance
Stepwise intensification once targets are missed despite adherence, per the guideline the clinic selects.source: clinic guideline
Context used
- Patient58 years, type 2 diabetes, no documented kidney impairment
- HistoryTwo prior HbA1c values trending upward over 9 months
- SettingPrimary-care clinic; follow-up by coordinator available
Would change the suggestion
- IfAdherence is confirmed as good: intensifying therapy becomes the leading option
- IfKidney function is reduced: drug choice and dose need review first
- IfSymptoms of hyperglycaemia are present: faster escalation is discussed
Synthetic example with illustrative values. Decision support only: the clinician decides.
How it thinks
Six steps, each one visible and open to challenge
Select a step to see what it does and how a clinician can push back on it.
Gather the record
Pulls the relevant facts from the patient's record, each tied to its source file, and notes what is missing.
You can challenge: any fact, with one click to its source.
List the options
Lays out the plausible diagnoses or actions, including the less obvious ones, the way a clinician builds a differential.
You can challenge: add or remove an option.
Weigh the evidence
For and against each option, with the strength of each piece of evidence stated, and probabilities that reflect the uncertainty.
You can challenge: the weight given to any item.
Check against itself
Looks for what would overturn its own conclusion, and for anything in the record that contradicts it.
You can challenge: run it again with a different assumption.
Write the report
Delivers the mini report: conclusion, why, probabilities, evidence, context and what would change the answer.
You decide: the clinician owns every decision.
The difference
A black-box answer versus a reasoned report
"Increase the metformin dose."
No reasoning shown, no evidence, no sense of how sure it is. You must either trust it or redo the work.
- WhyRise is modest, dose has headroom, adherence unconfirmed
- How sure54 % optimise, 31 % add agent, 15 % lifestyle recheck
- EvidenceLinked to record #R-204 and #R-198
- Would changeConfirmed adherence, reduced kidney function
Synthetic example.
Built on how clinicians and researchers think
Three habits we built in
Hypothesis first
Start from the plausible explanations and test each against the facts, instead of pattern-matching to one answer.
Calibrated uncertainty
Say how sure it is, and why, so a moderate-confidence suggestion is never presented as settled.
Open to refutation
Every step is exposed so a clinician can challenge it, the way a good colleague would.
Shishyan Thinking supports clinical decisions. It does not replace a clinician's judgement, and it does not talk to patients.
Part of one system
Reasoning is only as good as the record behind it
Early access
Decide with reasoning you can see
Request early access above. Our team contacts you and onboards you.
Request early access