ESSENTIAL KNOWLEDGE · NO CODING REQUIRED
A friendly field guide for researchers who want the useful ideas—not a computer-science degree.
THE BIG PICTURE
The model generates. The application gives you a place to work. Tools add capabilities. Skills make work repeatable. Agents coordinate steps.
SEVEN SHORT STOPS
MEET AI
AI can summarise, compare, draft and transform information. It does not independently know whether a claim is true. Treat it as a capable assistant whose work still needs evidence and judgement.
Fluency ≠ truthMAP THE ECOSYSTEM
The model is the engine. An application is the vehicle. Tools give it capabilities. Skills provide reusable ways of working. Agents coordinate multiple steps.
Choose the smallest thing that solves the taskINSIDE A CHAT
Your instructions, conversation, files and sources form the current context. Tokens measure how much text is processed. Clear, relevant context beats a giant information dump.
Goal + context + sources + constraints + output + checksGET STARTED
Begin with ChatGPT on the web or desktop. Use Codex when you want an agent to create files, analyse data, build a site or work with code. Start with a small, reviewable task.
Ask for an explanation before a changeRESEARCH JOURNEY
Explore questions, search and read literature, organise notes, analyse data, generate figures, revise prose and prepare presentations—without outsourcing scientific decisions.
AI assists the stage; the researcher owns the claimWORK WELL
Break complex work into stages. Ask for assumptions. Compare outputs with original sources. Record important prompts, model or application, date and human changes.
Verification is part of the workflowBUILD & EXTEND
When a task repeats, shape it into a skill. Use connectors or MCP when authorised external data is needed. Automate only after the manual workflow is trustworthy.
Reliable first, automatic secondVISUAL EXPLAINER
The AI can work only with what is available in its current context. Pack what matters; leave out noise and sensitive material.
AI ACROSS A RESEARCH JOURNEY
AI can assist every stage. The researcher still owns the question, method, interpretation and claim.
SEARCHABLE GLOSSARY
Computer systems that perform tasks associated with human intelligence—without thinking exactly like a person.
AI that creates new text, images, audio or code from patterns learned during training.
A large language model: the underlying engine that predicts and generates language.
A small unit of text processed by a model. Your prompt, files and answer all use tokens.
The information currently available to the model: instructions, messages, files and tool results.
Your request to the AI. Useful prompts clarify the goal, context, sources, constraints and output.
A complete product such as ChatGPT, Elicit or NotebookLM.
A capability an AI can use—such as web search, Python, file reading or image generation.
A reusable workflow that tells an AI how to complete a recurring task and check its work.
AI that can plan steps, use tools and work toward a goal while remaining under human direction.
A standard way for AI systems to connect with external tools and authorised data.
Retrieval-augmented generation: finding relevant sources first, then using them to support an answer.
A fluent but unsupported or incorrect AI response. Plausible wording is not evidence.
Connecting an AI response to supplied documents, trusted databases or verified web sources.
Able to work across more than one medium, such as text, images, audio and files.
YOUR FIRST 15 MINUTES
CHATGPT · SIMPLEST ROUTE
CODEX · NO-CODE FRIENDLY START
A PROMPT RECIPE WORTH KEEPING
ASK THIS GUIDE · NO AI TOKEN USED
This searches a curated answer library stored inside the page. It does not call an AI model, consume API tokens or send your question elsewhere.
AI is a broad term for computer systems that perform tasks associated with human intelligence. Generative AI creates new text, images, audio or code by learning patterns—not by understanding the world exactly as a person does.