Writing prompts that get reliable, useful results from language models.
What actually changes a model's output, in order of effect — and the habits that waste effort.
Making a model return data your code can rely on, every time.
Making a model work through a problem, and the techniques that genuinely help.
Treating a prompt as code: tests, versions, review and a rollback.
Defending a prompt-driven system against the text it reads, and against its own users.
The two most common production tasks, done reliably.
Getting useful code out of a model, and reviewing what comes back.
The handful of arrangements that most real features are built from.
35 in this topic
The parts of a prompt that reliably produce good results, with a template you can adapt for almost any task.
Prompting 1 min read 9 Nov 2025 For people
Why most weak outputs come from missing context, and what background information to include in a prompt.
Prompting 2 min read 8 Nov 2025 For people
Telling a model to act as an expert or a persona can shape tone and focus — but it isn't a substitute for clear instructions.
Prompting 1 min read 7 Nov 2025 For people
Precise, positive, prioritised instructions: how to phrase what you want so models follow it consistently.
Prompting 1 min read 6 Nov 2025 For people
How clearly marking documents, examples and instructions with tags or headings improves accuracy and safety.
Prompting 1 min read 5 Nov 2025 For people
When asking a model to reason before answering improves results, how to ask for it, and how to keep the final answer clean.
Prompting 1 min read 4 Nov 2025 For people
Verification prompts, second passes and critique loops that catch errors before you see them.
Prompting 1 min read 3 Nov 2025 For people
Splitting complex work into a sequence of focused prompts, where each step's output feeds the next.
Prompting 1 min read 2 Nov 2025 For people
How to get responses of the right length — no rambling, no truncation — with explicit limits and structure.
Prompting 1 min read 1 Nov 2025 For people
How to get writing that sounds right for the audience — from formal to friendly — including matching a house style.
Prompting 1 min read 31 Oct 2025 For people
How to express what a model must avoid, why positive framing usually works better, and when hard constraints need code.
Prompting 1 min read 30 Oct 2025 For people
Pulling structured fields from invoices, emails and documents reliably with clear schemas, examples and validation.
Prompting 1 min read 29 Oct 2025 For people
Using language models as classifiers: defining categories, handling ambiguity, getting confidence and evaluating accuracy.
Prompting 1 min read 28 Oct 2025 For people
How to get correct, maintainable code from AI: context, specifications, tests and review.
Prompting 1 min read 27 Oct 2025 For people
Turning transcripts into useful summaries with decisions, action items and owners — and avoiding invented commitments.
Prompting 1 min read 26 Oct 2025 For people
Treating prompt development like engineering: test cases, one change at a time, and records of what worked.
Prompting 1 min read 25 Oct 2025 For people
How to get accurate answers from long contracts, reports and transcripts: document placement, quoting and section-by-section work.
Prompting 1 min read 24 Oct 2025 For people
Getting the best from a conversation with a model: building context, correcting course and knowing when to start fresh.
Prompting 1 min read 23 Oct 2025 For people
How to ask good questions about images, charts, screenshots and documents, and how to check the answers.
Prompting 1 min read 22 Oct 2025 For people
Building reusable prompts with placeholders for repeated tasks, and keeping templates safe and maintainable.
Prompting 1 min read 21 Oct 2025 For people