Skip to content

Instantly share code, notes, and snippets.

@bishopb
Created April 27, 2026 18:59
Show Gist options
  • Select an option

  • Save bishopb/196f40ce4eb7451f618648cac25a5b58 to your computer and use it in GitHub Desktop.

Select an option

Save bishopb/196f40ce4eb7451f618648cac25a5b58 to your computer and use it in GitHub Desktop.

You are creating a Reasoning Framework, a set of guidance that will shape how the Bioscope AI reasons about clinical questions in your area of expertise. Each section below is a "frame", which is just a single paragraph of guidance about either a specific clinical situation or a general instruction for how the AI should communicate. You can have as many frames as you need. Each frame is just a paragraph answering a few key questions. Write naturally, as if you were explaining your thinking to a colleague.

Questions each frame should answer:

  1. When this applies: Describe the situation, clinical context, or condition under which this guidance should be activated. If this guidance applies to all conversations (such as communication style or formatting preferences), say so here.
  2. The usual approach: What is the standard or textbook practice for this situation? What would most clinicians do by default? If this is a communication or style directive, describe how the AI typically writes.
  3. What to consider instead: Your guidance. What should the Bioscope AI weigh, prioritize, or do differently? Write as if you're coaching a resident — be as specific or as general as you need to be.
  4. Based on: What evidence or experience informs this guidance? Cite studies, guidelines, institutional data, or clinical experience. This helps clinicians who read the AI's response understand the depth of your reasoning.
  5. Strength: How strongly do you recommend this over the usual approach on a 1 to 5 scale, where "5" means absolutely must be followed and "1" means should be considered.

Here are some examples:

APOE-guided lipid therapy

Patient with hyperlipidemia where statin therapy is being considered, particularly when APOE genotype is known or obtainable. Statins are the go-to response. High-intensity statin therapy is standard first-line for significantly elevated LDL. Begin with lifestyle modifications and supplement evaluation — red yeast rice, plant sterols, omega-3 — before moving to pharmacotherapy. When pharmacotherapy is appropriate, consider APOE genotype as part of a precision therapy approach. APOE4 carriers may have reduced LDL response to statins and should be evaluated for PCSK9 inhibitor candidacy earlier in the treatment sequence than conventional guidelines suggest. Based on 2023 Kathiresan et al. meta-analysis showing 23% reduced LDL response to atorvastatin in APOE4 carriers. Institutional outcomes data from PCSK9 inhibitor use in this subpopulation, 2022–2024, showing favorable lipid response and tolerability. Strength: 3

Communication style

Apply this to all conversations. The user is a busy physician. AI output phrasing is typically academic, dense, and hedging, with long paragraphs and frequent qualifications. Write in a style that supports rapid comprehension. Tone is direct, objective, and authoritative. Voice is active and front-loaded. Diction is simple, concrete, and specific. Register is professional and conversational. Keep paragraphs to 1–3 sentences. Use bulleted lists for 3 or more items. Use specific header levels to categorize. Use bolding to highlight key phrases, but sparingly. Strength: 5

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment