THE CRAFT
The Craft
How to actually work with these tools — prompting, agents, automations, and the honest cases where reaching for AI costs you.
PAGES51
WORDS83,657
SOURCES77
READ TIME189 min
CHECKED25 AUG 26
How to work with these tools. Every prompt and method carries the date it was last tested.
◈ WHY THIS SECTION EXISTS
Every other part of this site tells you what is happening. This one is about what to do with the tools yourself — because the most common outcome of learning how AI works is not avoiding it. It is using it badly, confidently, and without noticing.
The three things this section teaches
Context beats phrasing
The industry called it "prompt engineering" and the name did damage, because it implied secret words. A model fills every gap you leave with the average of everything it has seen — so the work is supplying the specifics only you have, not hunting for magic formulas.
Techniques tied to one model's quirks expire at the next release. This does not.
Keep the judgement, hand over the labour
Decide before you start which part of the work is the part you are for — the position, the recommendation, the thing you would defend in a meeting.
If you cannot explain why an output is right, you have not done the work. You have forwarded it.
Verification is the whole job now
These systems produce confident, fluent text whether or not the claim underneath is true, and fluency carries no information about accuracy. That single fact converts "using AI well" from a prompting skill into a checking discipline.
Where to start
If you read one page here, read how to ask AI well — seven habits that survive model releases. Then when not to use AI, which covers where these tools make the outcome worse, and using AI at work if a job is involved.
◈ THE ADMISSION THIS SECTION IS BUILT ON
The site’s editor works in generative AI and uses these tools on parts of this site, and says so on every relevant page. That is not a disclaimer bolted on at the end — it is why this section can be specific.
The people best placed to say where a tool fails are the ones using it constantly, and almost none of them write it down.
What is here
01 · 34 MINThe Prompt VaultThirty-nine production prompts for image, video, LLM and pre-production work — four with a test date, the rest reviewed.NEW · 11 MIN · 23 SOURCESHow to Test Whether a Prompt Change Actually HelpedA simple way to test whether a prompt change helped: fixed test inputs, one change at a time, several runs each, and “better” defined before you look..NEW · 12 MIN · 28 SOURCESPrompting Myths: Advice That Stopped Being True (or Never Was)Common prompting rules, from “think step by step” to “8K masterpiece”, checked against what OpenAI, Anthropic, Google and image vendors actually publish..NEW · 14 MIN · 22 SOURCESFree Alternatives: Run AI on Your Own ComputerHow to run AI free on your own computer with Ollama or LM Studio: the memory you need, the steps, which models to try, and what each licence allows..NEW · 5 MINWhat Generation Actually CostsFour pricing models, none comparable by looking. How to budget a real job.NEW · 6 MIN · 1 SOURCESGetting Text Right Inside an ImageIt mostly works now. Where it still fails is specific enough to plan around.NEW · 6 MINPrompting Music and VoiceSound happens over time. Most audio prompts describe it like a picture.NEW · 7 MIN · 3 SOURCESPrompt Injection: Why There Is No FixEvery other injection attack has a fix. This one does not.NEW · 8 MIN · 3 SOURCESPhotorealistic, Hyperrealistic, Cinematic, AnamorphicFive words used as if they meant the same thing. Two point opposite ways.NEW · 8 MIN · 3 SOURCESPlan It Before You Generate ItBlock the camera in 3D, lock the faces, then generate.02 · 18 MIN · 6 SOURCESModel GuidesAround twenty models and studios with their optimal prompt format — ChatGPT Images 2.5, Seedance, Kling, FLUX,…03 · 13 MIN · 2 SOURCESPrompting Cheat SheetsFour reference sheets — image, video, LLM and AI studios.04 · 1 MIN · 6 SOURCESNegative Prompt LibraryCopy-ready negative prompts by use case.05 · 6 MIN · 1 SOURCESPrompting AgentsBriefing something that acts: the specification, the six ways agents fail, the dry run, and designing work so you can…06 · 3 MIN · 2 SOURCESContext EngineeringPrompting is now the small part of the job.07 · 3 MIN · 2 SOURCESAgent GuardrailsFour lines that belong in every agent prompt.08 · 3 MIN · 2 SOURCESA Security Checklist for AI-Built SoftwareSecurity checks for software you did not fully write, in three tiers by what a mistake costs, plus three…09 · 3 MIN · 2 SOURCESLoops That Improve and Loops That DegradeSome iteration converges and some polishes its own output while appearing to improve.10 · 3 MIN · 2 SOURCESWhat You Owe a Client Whose System You AutomatedWhat you owe a client after automating their work: a handover document, disclosure, an off switch, a visible…14 SHEETS · VENDOR-SOURCEDModel Prompting Cheat SheetsOne sheet per model — the vendor's own formula, whether it takes a negative prompt, and a checklist.NEW · 9 MIN · 10 SOURCESWhat the Vendors Actually SayThe prompt formula each vendor publishes for its own model — image, video, text and audio — and the negative-prompt table nobody prints.11 · 4 MIN · 1 SOURCESBad, Good, GreatThe same goal written three ways.12 · 5 MINAutomationsChains that run without you.13 · 3 MIN · 1 SOURCESThe Agent BriefA copyable structure for briefing an agent — objective, scope, done-condition, stop-conditions, evidence — with why…14 · 3 MIN · 1 SOURCESSkills Are a Supply ChainInstalling a skill means loading someone else’s instructions into your agent’s context.15 · 3 MIN · 1 SOURCESWhat You Are Actually BurningYou are billed for everything the model reads, re-sent every turn.16 · 2 MIN · 1 SOURCESMulti-Agent: When It Helps and When It Does NotWhy chaining agents cuts reliability — compounding error, lossy handoffs, failures you cannot trace — and the cases…17 · 4 MINReverse PromptingFeed any AI output back to a model and reconstruct the prompt behind it — then improve on it.18 · 3 MINReusable Context Beats Reusable PromptsPrompts encode how to ask a particular model and expire with it.19 · 3 MINCamera GrammarTwenty-four shot types in five families — angle, distance, movement, focus, composition — what each reads as, and why…20 · 2 MINDiagnosing a Run That Went WrongThe agent finished, the output is wrong, and you have a transcript.21 · 2 MINThe Review Loop, In OrderFrame, then facts, then structure, then language.22 · 2 MINPrompt, Context, Skill, or Fine-TuneFour ways to shape model behaviour at wildly different costs, and how to choose.23 · 2 MINSkills, Repos and MCPWhat these three actually are, explained without assuming you write code.24 · 2 MINPermission Tiers for AgentsFive permission levels for agents, set per task by how hard an action is to undo, not by how far you trust the… ◈ THE OTHER SIDE
Most pages here teach what to do. When not to use AI covers the other side — where these tools are worse than doing it yourself, and how to tell in advance.