Toolkit
Every AI platform, prompt template, and technical resource I recommend -- organized and ready to use. Browse by type or search for exactly what you need.
My AI Toolkit: What I Actually Use Every Day
The AI tools I actually use as a physician and researcher, organized by task. Updated monthly.
How We Use NotebookLM: A Radiation Oncologist's Workflow
First-person account of using NotebookLM for disease-site knowledge management, grounded Q&A, and AI-powered podcasts.
All Resources
When Should I NOT Use ChatGPT? Matching AI Tools to Tasks
General-purpose vs specialized AI tools. Know which tool to reach for.
Prompt Engineering for Clinicians
Practical framework (CREF method) for writing effective clinical prompts.
The Challenge Circle
Don't just ask for agreement. Use AI to stress-test your ideas.
ChatGPT by OpenAI
Integrated with Apple Shortcuts for system-wide voice workflows. Strong general-purpose model.
Claude by Anthropic
My go-to for writing code, working with data, and complex reasoning tasks.
Gemini by Google
For general knowledge queries and quick lookups. Strong web search integration.
Perplexity
AI-powered search engine that cites its sources. Perfect for fact-checking and quick research questions.
Microsoft Copilot
AI assistant integrated into Microsoft 365. Useful for email drafting, PowerPoint creation, and Excel analysis.
Superwhisper
Dictation → LLM → text. Game-changer for computer interaction.
Whisper Flow
Alternative to SuperWhisper for voice-to-text dictation workflows.
Granola
AI-powered meeting notes that let you stay present. Records, transcribes, and summarizes automatically.
Notion
Second brain and productivity manager. My central hub for everything.
Doximity GPT
AI assistant for physicians with access to medical literature and clinical synthesis.
ElevenLabs
AI text-to-voice. Convert text to natural-sounding audio.
Claude Code
Anthropic's official CLI coding agent for actual software development, debugging, and code generation from the terminal.
Cursor
AI-powered code editor built on VS Code with inline code generation, chat, and codebase understanding.
ScholarAI
AI research assistant for literature search and synthesis, built for academic workflows.
Mastering Prompt Engineering (Jeff Su)
Clear, practical video tutorial on prompt engineering techniques.
Prompt Engineering for Developers (DeepLearning.AI)
Short course co-taught by Andrew Ng and OpenAI. Slightly technical but extremely practical.
Anthropic's Interactive Prompt Engineering Tutorial
Anthropic's official hands-on tutorial for engineering optimal prompts with Claude.
Anthropic Academy
Anthropic's official learning hub with courses on AI fluency, API development, Model Context Protocol, and Claude Code.
Prompt Design Subreddit
Community-driven collection of prompts, tips, and shared experiences.
Awesome ChatGPT Prompts (GitHub)
Open-source collection of useful prompt templates for various use cases.
Google Colab
Free Jupyter notebooks with GPU access. Run Python and ML experiments without any local setup.
Hugging Face
The hub for open-source AI models, datasets, and demos.
GitHub
Where code lives. Search for repos, tools, and projects related to problems you're solving.
Apple Shortcuts + AI Integration
System-wide AI workflows on iOS and Mac using Apple Shortcuts with LLM integrations.
Gamma
AI-powered presentations. Generate slides from text prompts--great for quick decks.
Excalidraw
Simple whiteboard tool for quick sketches and diagrams. Fast, collaborative, and free.
Nano Banana Pro
Phenomenal AI image generation. Needs extensive prompting but produces professional medical and scientific visuals.
The Jargon Translator
Converts dense medical notes or radiology reports into patient-friendly language.
Letter of Medical Necessity
Generates a strong appeal letter citing specific guidelines.
The Differential Diagnosis Challenger
Helps broaden the differential diagnosis and avoid anchoring bias.
The 1-Minute Paper Summary
Quickly extracts the key data points from a PDF.
Abstract Shrinker
Reduces word count while preserving scientific density.
Data to Graph (Python)
Generates code to visualize data.
Meeting Note Polish
Turns messy typing into clean minutes.
The Board Examiner
Creates study materials from lecture notes.
The Diplomatic Email
Helps say 'no' professionally.
Consent Form Simplifier
Makes legal documents readable.
Specific Aims Refiner
Provides high-level feedback on grant structure.
Presentation Script
Writes the speech for your slides.
Excel Wizard
Generates complex Excel/Sheets formulas.
Dietary Advice Generator
Generates personalized supportive care materials.
LOR Drafter
Drafts LORs based on specific traits.
Large Language Models Encode Clinical Knowledge
The landmark paper introducing Med-PaLM, the first AI to reach passing score on the USMLE style questions.
Capabilities of GPT-4 on Medical Challenge Problems
Demonstrated that GPT-4 could pass medical licensing exams without specialized training, highlighting its reasoning capabilities.
Comparing Physician and AI Chatbot Responses to Patient Questions
Famous study showing that evaluators preferred chatbot responses to physician responses 79 percent of the time, rating them as higher quality and more empathetic.
Performance of ChatGPT on the Radiation Oncology In-Training Exam
Assessed ChatGPT's ability to answer specialized radiation oncology board questions, identifying strengths in general concepts and weaknesses in physics/dosing.
Large Language Models in Medicine: The Risk of Hallucinations
A comprehensive survey of hallucination types in LLMs and their specific implications for medical safety.
Attention Is All You Need
The seminal paper that introduced the 'Transformer' architecture, which forms the basis of all modern LLMs (GPT, BERT, Claude).
Constitutional AI: Harmlessness from AI Feedback
The foundational paper from Anthropic detailing how to train AI models (like Claude) to be helpful and harmless using a set of principles.
Performance of GPT-4 in Radiology Board Examinations
Evaluated GPT-4's performance on radiology board exams, showing high accuracy in text-based questions but limitations in image interpretation.
Coding Bias: AI in Healthcare
A critical study revealing racial bias in a widely used commercial algorithm for guiding health decisions.
Scaling Laws for Neural Language Models
The paper that proved 'bigger is better' - showing that model performance improves predictably with more data and compute.
Want more?
Subscribe to Beam Notes for updates on AI in medicine.
Subscribe to Newsletter