From your first question to prompt engineering and automated prompting.
Write the question and the decision it serves first, decide which sources count as evidence, then give the AI narrow jobs such as searching, summarizing and listing disagreements, and check every claim you…
4 min read
Take a prompt you have already refined, move its instructions into a SKILL.md file with a clear name and a description of when to use it, and add a worked example and the edge cases you have met.
4 min read
Hide your own opinion, ask for evidence on both sides, and give the AI a concrete job such as finding flaws or scoring against criteria.
4 min read
For everyday requests, yes: a short prompt you write yourself, refined over a few turns, is still the fastest route.
4 min read
Context engineering is the work of deciding everything a language model sees before it answers, including instructions, retrieved documents, tool definitions, memory and conversation history, and assembling…
4 min read
An automatic prompt optimizer treats the prompt as something to search for: it tries many versions of the instructions and examples, scores each one against labeled examples with a metric you define, and…
4 min read
Give the chatbot a different job: instead of solving your problem, it asks you questions, one at a time, that help you lay out the facts, weigh the options and reach your own decision.
4 min read
Prompt a reasoning model like you would brief a capable new colleague: give it all the context up front, state exactly what result you want and how you'll judge it, and leave the method to the model.
3 min read
A prompt improver is a tool that takes a prompt you already have and asks a model to rewrite it with better structure, examples and reasoning steps.
4 min read
Ask about the answer you just got, not about the topic in general: why it is true, what evidence supports it, what else could explain it, and what would change your mind.
4 min read
Prompt chaining means breaking a large job into a sequence of smaller prompts, where the output of each step becomes the input of the next.
4 min read
Meta-prompting means asking an AI model to write the prompt you will then use, instead of drafting it yourself.
4 min read
Ask for a draft, then ask the model to grade that draft against a short list of criteria you define, quoting the weak parts, and finally to rewrite it fixing only what the critique found.
3 min read
A team prompt library is a shared, searchable collection of tested prompts, each stored with its purpose, the prompt text with fill-in blanks, a sample of good output, an owner and the date it was last tested.
4 min read
A system prompt is the standing brief you give Claude before the conversation starts: who it is, what job it does, which rules it follows and which documents it should use.
4 min read
Start by asking the AI which questions an expert would ask about your situation, answer those questions with your own details, and only then ask for advice.
4 min read
Keep a simple spreadsheet or notes file with one row per prompt version, recording the task, the exact prompt, a short note on the result, a rating and a version number.
4 min read
Ask ChatGPT to argue both sides of your idea on purpose: first to state the best possible case for it, then to attack that case as a skeptical expert would.
4 min read
Show it how you write instead of only telling it. Paste two or three short samples of your own writing, add a few plain sentences describing your style, then ask for the new text and correct the first draft…
3 min read
A reusable prompt template is a prompt with fixed wording and labeled blanks, such as role, audience, task, format and input, that you fill in fresh for each job.
4 min read
Prompt engineering is the work of designing the text you give a language model so that it performs a task as well as possible, then testing and refining that text.
4 min read
Chain-of-thought prompting means getting a language model to write out the intermediate steps of a problem before it gives the final answer, either by showing it worked examples that include those steps or by…
4 min read
Sharpen the idea in passes before you ask: say who it is for, what result you want, what limits apply, and what shape the answer should take.
3 min read
Edit mode takes a piece of existing text plus an instruction, such as "make this friendlier," and returns a revised version; insert mode takes the text before and after a gap and writes what belongs in…
4 min read
Zero-shot prompting means asking a language model to do a task with only an instruction and no worked examples in the prompt.
3 min read
Write to an instruction-following model the way you would brief a capable assistant: start with the task as a direct request, then say who the result is for, how long it should be and what shape it should…
4 min read
Show GPT-3 the exact format two or three times, in identical worked examples separated by clear delimiters, and end each example with the same end marker.
4 min read
Temperature and top-p control how GPT-3 picks each next word from its list of likely candidates.
4 min read
Few-shot prompting means putting a handful of solved examples of a task directly in the prompt, so a language model like GPT-3 can see the pattern and apply it to a new case.
4 min read
A good GPT-3 prompt sets up a clear pattern for the model to continue: a short description of the task, a few worked examples in a consistent format, and an open slot where the answer should go.
4 min read
Sign up for the weekly newsletter and receive the free guide: what changed in AI this week, one lesson, one prompt to try. Unsubscribe in one click.