AI tools and prompting guide.
Practical guides for biologists on using AI tools for literature review, data analysis, coding assistance, and writing workflows.
Our mission is to understand life and champion open science by sharing that knowledge with the world. We use AI to remove barriers to that mission, to code faster, write clearer, and think bigger. These powerful tools demand equally rigorous oversight. We use AI to expand our minds, not replace them, demanding 100% verification so our work is transparent, reproducible, and stands on a bedrock of accuracy.
Frontier models, the engines
These are not simple chatbots. They are reasoning engines. With the right prompting strategy (context, persona, constraints), they can draft entire manuscripts, synthesize vast literature, and code complex pipelines. Used lazily, the output is garbage. Driven like a Ferrari, you win.
OpenAI ChatGPT The original heavy hitter, strongest on complex reasoning in its top-tier models.
The "o1" family is currently the king of complex reasoning. If you are using the free version (GPT-3.5 or 4o-mini) for science, you are wasting your time. Pay for the real tools.
Anthropic Claude Currently the best prose writer of the bunch, and strong on long-context inputs.
Sounds less robotic than the others and handles massive text inputs (whole books) beautifully. If you need to draft a paper, start here.
Google Gemini Fully integrated into the Google ecosystem, with a massive context window.
Brilliant at multimodal tasks, but can be confidently wrong. Do not let it think for you. Always verify the output. Currently my go-to for many tasks.
The power of these tools is highly dependent on your prompting. See the guide to prompting, especially the mega-prompts section, where the real capability lives.
Prompt engineering, tools to build tools
If you are building complex workflows (mega-prompts), you cannot just type into a chat window. You need environments that let you test, iterate, and save your source code.
Anthropic Anthropic Workbench Best place to build prompts, with a "Generate a Prompt" feature that writes structured mega-prompts for you.
Essential for system-prompt testing. Describe a task and it writes a structured mega-prompt you can iterate on.
OpenAI OpenAI Playground Lets you edit system and user messages separately, critical for keeping the model in character.
Stop using the normal ChatGPT interface for complex tasks. The Playground separates the System message from the User message, which is critical for getting the AI to stay in role.
Open source Fabric A library of "Patterns" (mega-prompts) for analyzing life, science, and security.
Open-source project by Daniel Miessler. A goldmine for seeing how experts structure their requests.
Jina PromptPerfect Rewrites your rough prompt into a detailed, optimized version.
Good for learning how to be more specific, and as a sanity check on your own prompts.
Third-party TypingMind A better UI for ChatGPT and Claude, with a prompt library you can call with a slash command.
Save your favorite mega-prompts and trigger them with commands like /review.
Literature search and discovery
Google NotebookLM AI research assistant grounded in your uploaded PDFs and sources.
Answers questions and summarizes strictly from the material you provide, and can generate podcast-style audio summaries of your documents.
Elicit Elicit Automates literature reviews, with summaries in a table format.
Finds relevant papers and summarizes key takeaways in a side-by-side table that is great for scoping a new topic.
ResearchRabbit ResearchRabbit Visualizes citation networks to help you discover papers near your seed collection.
Starts from a small set of papers you already trust and expands outward along citation, co-author, and similarity links.
Consensus Consensus AI-powered search that extracts answers directly from peer-reviewed literature.
Aggregates findings across many papers, which is useful when you want to see the weight of evidence for a specific claim.
Scite Scite Classifies citations as supporting, mentioning, or contrasting with the cited claim.
Invaluable when you want to know whether a paper is widely backed, quietly contested, or routinely contradicted.
Google Google Scholar AI Gemini chat with access to Google Scholar.
Great for finding papers that speak to a complex idea or question you cannot yet phrase as a single keyword query.
Google Google Scholar PDF Reader Adds an AI sidebar to online PDFs, with outlines and summaries while you read.
Sits next to a paper in your browser and gives you inline outlines, summaries, and jump points.
Connected Papers Connected Papers Graph of similar papers based on co-citation and bibliographic coupling.
Good for seeing the shape of a literature at a glance, especially when starting a new topic.
Scholarcy Scholarcy Breaks down complex papers into digestible flashcards.
Highlights key findings and methodology, which is useful for keeping a running reading log.
Writing and editing
Paperpal Paperpal AI writing assistant trained on academic manuscripts.
Checks grammar and style against journal standards, not generic English.
Writefull Writefull Language feedback using a corpus of published scientific papers.
Compares your phrasing against how published scientists actually write, which is a better rubric than generic grammar rules.
DeepL DeepL Highly accurate AI translation.
Useful for clarifying complex phrasing, and for working with collaborators across languages.
Grammarly Grammarly Standard tool for catching typos, tone, and basic grammatical errors.
Good as a final sweep before a draft goes out, not as a substitute for careful reading.
Data analysis and coding
GitHub GitHub Copilot AI pair programmer that integrates with RStudio and VS Code.
Autocompletes code in context. Best treated as a fast junior collaborator whose work you still need to read.
Julius Julius AI Data analysis assistant for plain-English queries on uploaded datasets.
Upload a dataset, ask questions in natural language, and get plots and summary tables.
RTutor RTutor Generates R code and plots from natural language.
Great for syntax checks and for quickly prototyping a plot before you refine it by hand.
Figures and visualization
BioRender BioRender Industry standard for professional biological diagrams and pathways.
Prebuilt icons for cells, organisms, and pathways, with enough flexibility to make publication-ready figures.
Adobe Adobe Firefly Generative tool integrated into Adobe apps, for vector graphics from text.
Useful for conceptual art and cover figures, especially when you need vector output.
Generative art Midjourney / DALL-E Generative AI for conceptual cover art or presentation backgrounds.
Strong for illustrative work at the edges of a talk or paper. Do not treat the outputs as factual diagrams.
Teaching and presentations
Gamma Gamma Generates polished slide decks from a simple text outline.
Good starting point for a deck you will then tune by hand.
Curipod Curipod Interactive presentation tool with AI-generated polls and questions.
Useful for live classrooms where you want students responding to prompts as you teach.
Quizlet Quizlet Automatically generates flashcards and practice tests from lecture notes.
Best for student self-quizzing, especially on vocabulary-heavy content.
Productivity and lab management
Otter.ai Otter.ai Meeting assistant that records audio and generates summaries.
Useful for lab meetings, but run it only with the group's consent and with a clear retention policy.
Notion Notion Workspace with AI for organizing lab wikis and summarizing protocols.
Good central home for protocols, onboarding docs, and running lab wikis.