How to use AI, actually.
Companies will happily charge you $49 a month to learn how to use a tool that teaches anything for free. This guide skips the paywall. It takes the same eight-skill framework those certificates teach — prompting, planning, research, writing, content, data, building, shipping — and develops every one of them into working knowledge: real prompts, real before-and-afters, and the hacks that make the difference between a toy and a tool.
The single most important idea in the whole guide fits in one sentence: an AI doesn't read your mind, it completes your pattern. Give it a vague pattern, you get a vague completion. Give it role, context, examples, and constraints, and it starts to feel like a colleague.
The 5-part prompt formula
Almost every disappointment with AI traces back to a missing ingredient in the prompt. Almost every “wow” moment traces back to all five being present. Memorize this and you're ahead of most paid-course graduates.
↳ Generic because the prompt was generic. No recipient, no purpose, no outcome — so the model filled the blanks with the most average email ever written.
↳ Same model, same task. The only difference is the prompt. That's the whole game — and everything below is variations on it.
Try it: the Prompt Builder
Fill in the five parts and watch a working prompt assemble itself. Then copy it into any AI — Gemini, ChatGPT, Claude, whichever you use. The techniques in this guide are model-agnostic.
AI Fundamentals
You don't need to know how a large language model is built to use one brilliantly — but three facts about how it works will save you from 90% of bad outputs.
1. It predicts, it doesn't look up
An AI generates the most plausible next words given your prompt and its training. It has no live database of facts unless it's connected to search. Plausible ≠ true — which is why verification (Chapter 3, and Responsible Use) is a skill, not an afterthought.
2. Context is its whole world
Within a conversation, the model can only “see” what's in the chat (its context window). It doesn't remember yesterday's separate chat, your files, or your intentions. If it matters, paste it in. Front-load the important stuff — models attend most to the start and end of long inputs.
3. It's a mirror with range
Ask like a beginner, get beginner answers. Ask like an expert — precise terms, real constraints, examples of what good looks like — and the model's “expert mode” activates. The quality ceiling of the answer is usually set by the quality floor of the question.
The iteration loop — the skill underneath all skills
Nobody gets the perfect output on the first prompt, including the people who teach the paid courses. Professionals iterate: prompt → read critically → steer → repeat. Steering language that works:
- “Same content, half the length, punchier opening.” — surgical edits beat starting over.
- “You assumed X. Actually Y — redo it with that.” — correcting an assumption fixes a whole class of errors at once.
- “Give me 3 versions: safe, bold, and weird.” — range first, judgement second.
- “Before you answer, ask me the 3 questions that would most improve your answer.” — the single most underused move in prompting. The model knows what information it's missing; make it say so.
End every first prompt with: “Ask me clarifying questions first.”
For anything non-trivial, append: “Before you start, ask me up to 4 questions that would most change your approach. Then wait.” You answer, then it produces. This one habit converts one-shot guessing into a briefed collaboration — it's the closest thing to a free upgrade that exists in prompting.
Show, don't just tell — few-shot examples
Instead of describing the style you want for three sentences, paste two examples of it and say “match this pattern.” One example is a hint; two is a specification. Works for tone, formatting, code style, naming conventions, anything with a pattern.
Try it now (5 minutes)
Take any question you asked an AI this week that got a mediocre answer. Re-ask it with the 5-part formula plus “ask me clarifying questions first.” Compare the two answers side by side. That gap is your prompting skill, made visible.
Brainstorming & Planning
AI's superpower isn't knowing answers — it's generating options faster than your inner critic can kill them. Used right, it's a brainstorming partner that never gets tired, defensive, or attached to its first idea.
The diverge → converge method
Phase 1 — Diverge. Force quantity and range. Ask for ideas in deliberately different categories so the model can't give you 20 versions of the same thought:
I need ideas for [goal: e.g. a 30th birthday for a friend who hates big parties]. Give me 15 ideas, forced into these categories: - 5 that cost almost nothing - 5 that are experiences, not events - 5 that are slightly absurd but genuinely doable No explanations yet — one line each. Prioritise ideas that [constraint: work for a group of 6, on a weeknight].
Phase 2 — Converge. Now make the same conversation do the judging. Feed the shortlist back with criteria:
From the ideas above, I liked #2, #7, and #11. Score each 1–10 on: fits the person, effort for me, memorability, and risk of falling flat. Put it in a table, then recommend one and tell me the strongest argument AGAINST your recommendation.
↳ That last clause — “argue against your own recommendation” — is a hack in itself. Models default to agreeable. Explicitly requesting the counter-case is how you get honesty out of them.
From idea to plan: the backwards build
Ask for the plan built backwards from the outcome, with failure modes included:
Goal: [e.g. run a small community workshop in 6 weeks]. I have [time: ~4 hrs/week] and a budget of [$200]. Build the plan backwards from the day itself: 1. Week-by-week milestones, each with one concrete deliverable 2. For each week, the most likely way it goes wrong and a cheap prevention 3. A "minimum viable version" if I fall a week behind Keep it realistic, not motivational.
The pre-mortem
For any plan, ask: “It's six months later and this failed completely. Write the story of why — the 5 most plausible causes.” Models are far better at imagining concrete failure than at listing abstract “risks.” Then ask it to rebuild the plan so each failure story can't happen.
Stuck between options? Convene a panel: “Argue this decision three ways: as a cautious accountant, as an ambitious founder, and as my future self 10 years from now. Each gets 100 words, then they vote.” Different lenses surface trade-offs a single answer flattens.
Research & Insights
AI can compress a week of reading into an afternoon — if you treat it as a research assistant whose work you verify, not an oracle whose word you copy. This chapter is equal parts power and self-defence.
The research workflow that actually works
- Map the territory first. “What are the main schools of thought / competing explanations / key terms in [topic]? Give me the map before the details.” A map stops you asking narrow questions about a field you don't yet understand.
- Go deep with sources attached. Use a search-connected tool (Gemini Deep Research, ChatGPT search, Perplexity) and demand citations: “Every claim needs a source I can open. Prefer primary sources and recent ones; flag anything you couldn't verify.”
- Interrogate, don't just absorb. “What's the strongest criticism of this view? What would change an expert's mind? Where do sources disagree?”
- Verify the load-bearing facts yourself. Open the two or three sources your conclusion actually depends on. AI summaries of sources are usually good; AI-invented sources are a real failure mode (see Responsible Use).
↳ Note the moves: a fixed structure, evidence graded, omissions surfaced, and exact quotes demanded for numbers. “Summarise this” gets you a book report; this prompt gets you an analysis.
“Explain it at three altitudes”
For any hard topic: “Explain [X] three ways: to a smart 12-year-old, to a university student, and to a practitioner — then tell me what the expert version knows that the simple version hides.” The gaps between levels are where real understanding lives.
Turn AI into your tutor, not your answer key
Learning something? Say: “Teach me [topic] Socratically. Ask me one question at a time, correct my answers, and don't move on until I've got it. Start by finding out what I already know.” Being questioned beats being lectured — retention isn't close.
Models invent plausible-looking references — real journal names, fake article titles, dead links. Rule: if you can't open it, it doesn't exist. Ask for URLs and check the load-bearing ones. This is the #1 way AI research goes wrong in public.
Writing & Communicating
The highest-value everyday use of AI, full stop. Not because it writes for you — because it makes the distance between “rough thought in your head” and “clear thing on the page” almost zero.
Before / after: the difficult email
| What you type at 11pm | The prompt that fixes it |
|---|---|
| “per my last email the deadline was friday and the file still isnt here, this is the third time, we really need this sorted” | “Rewrite this as a professional email. Firm about the deadline, generous about the person — assume there's a reason I don't know. Give me two versions: gentle nudge and final notice. Keep my facts exactly as stated, invent nothing.” |
The four writing jobs AI does best
First drafts from bullets
Dump your messy points, voice notes transcript, or half-sentences in and say “turn this into a structured draft, my points in my order, flag anything unclear rather than guessing.” Your thinking, its typing.
The ruthless editor
“Edit for clarity only. Cut 30%. Flag every sentence where a reader could misunderstand me. Don't change my voice — here's a sample of how I write: [paste].” Voice samples are the difference between edited-you and generic-corporate.
Tone translation
Same message, different rooms: “Give me this announcement in three registers — for my team, for senior leadership, and for customers.” Also invaluable in reverse: paste a confusing message you received and ask “what is this person actually asking for, and what's the subtext?”
The blank-page killer
Staring at an empty doc is a prompting problem. “Give me 5 possible opening lines for [thing], ranging from safe to provocative.” You'll reject four and steal the fifth's angle. Momentum, not authorship, is what's being outsourced.
“Match my voice” — build a voice sample once, reuse forever
Paste 3–4 paragraphs of your own best writing and ask the AI to describe your style as rules (“short sentences, dry humour, no exclamation marks…”). Save that style card. Paste it into any future writing prompt and the output stops sounding like everyone's AI and starts sounding like you.
Here is a draft: [paste draft] And here is how I write at my best: [paste 1 paragraph of your own writing] Edit the draft to: 1. Match the voice of my sample 2. Cut anything that doesn't earn its place — I want it ~30% shorter 3. Keep every fact, name, and number exactly as written 4. List the 3 weakest sentences and why, BEFORE your edited version Do not add claims I didn't make.
Content Creation
Images, video, audio, social posts, presentations — AI collapsed the cost of a “first version” to nearly nothing. The skill now is direction: knowing what you want, describing it precisely, and curating hard.
Image prompting: it's art direction, not a search box
Image models (Gemini's image generation, DALL·E, Midjourney, Flux) fail the same way text models do — vague in, generic out. A strong image prompt has the same DNA as the 5-part formula: subject + medium/style + composition + lighting/mood + constraints.
- Negative prompts matter: “no text, no watermark, no extra fingers, no blur” — telling it what to exclude is half of directing.
- Iterate by edit, not re-roll: “Same image, move the light to evening and empty the counter.” Modern tools edit images conversationally; treat each generation as a draft.
- Reference images beat adjectives: upload a photo whose composition or palette you like and say “this layout, new subject: …”
Repurposing: make once, publish five ways
The highest-leverage content workflow: one substantial thing (a talk, a long post, a video transcript, a report) becomes many. One prompt:
Here is my [article / transcript / notes]: [paste] From ONLY the ideas in this source, produce: 1. A 150-word summary for a newsletter 2. Five short social posts, each built on a different single idea, none starting with the same word 3. Ten headline options for it, from straightforward to provocative 4. A script outline for a 3-minute video version 5. Three pull-quotes worth designing as graphics Do not introduce facts or claims that aren't in the source.
The anti-slop pass
AI content has tells: “delve,” “in today's fast-paced world,” everything in threes, relentless enthusiasm. Run this on any draft: “Rewrite removing AI clichés and filler. Vary sentence length wildly. Replace abstractions with one concrete detail. If a sentence could appear in any article on any topic, cut it.” Specificity is the whole difference between content and slop.
Don't ask AI to “make a presentation about X.” Ask it to first produce the narrative outline — claim per slide, evidence per claim — argue about that (cheap to change), and only then generate slides. Structure is the expensive part to fix later; bullet wording isn't.
Data Analysis
You don't need to be a data analyst to interrogate a spreadsheet — you need to know how to ask, and how to catch the model when it's confidently wrong about arithmetic (its most reliable character flaw).
Worked example: a small business's monthly sales
Say you paste this example table into an AI (Gemini in Sheets, ChatGPT, and Claude all accept files or pasted tables):
| Month | Orders | Revenue | Refunds |
|---|---|---|---|
| Jan | 412 | $18,540 | $610 |
| Feb | 388 | $17,220 | $590 |
| Mar | 501 | $24,180 | $1,340 |
| Apr | 536 | $25,910 | $1,520 |
| May | 489 | $22,760 | $980 |
Analyse the table I pasted. Rules: show your calculation for every number you state, and use a code/calculation tool if you have one rather than mental arithmetic. 1. Revenue per order by month — is it rising or falling, and from when? 2. Refunds as a % of revenue by month — flag any month that breaks the pattern 3. The 2 most plausible explanations for the March–April refund spike, and what extra data would distinguish between them 4. One thing this data CANNOT tell me, stated plainly
Describe the chart before you draw it
Ask: “What chart best shows the refund-rate break, and why?” Then have it build it. The story in the example data is the refund rate doubling in March:
“Show your working, then check it a different way”
For anything numeric: demand shown calculations, and finish with “Now verify your three key numbers using a different method and report any discrepancy.” Models are weak mental calculators but decent checkers — and tools with code execution (ask explicitly) eliminate most arithmetic errors entirely.
Before analysis, ask the AI to audit the data itself: “List duplicates, missing values, inconsistent formats, and outliers in what I pasted — don't analyse yet.” Ten seconds of cleaning prevents the classic chart that's precise, pretty, and wrong.
App Building
The newest skill in the framework, and the most startling: describing software in plain language and getting working software back. “Vibe coding” is real — with real limits this chapter won't hide from you.
What you can actually build today, as a non-programmer
Tools that fit in one page
Calculators, converters, quizzes, trackers, timers, generators, simple dashboards, landing pages, interactive explainers. Ask Gemini (Canvas), Claude, or ChatGPT to build it in a single HTML file and it will run in front of you.
Bigger, with an AI coding partner
Tools like AI Studio, Replit, Lovable, Bolt, and Cursor scaffold multi-file apps from descriptions. The pattern is the same — describe, run, react, refine — just with more moving parts to keep track of.
Worked example: from sentence to working app
Build me a single-page web app in one HTML file: a "focus timer" for deep work. - Big countdown display, presets for 25 / 50 / 90 minutes, plus a custom minutes input - Start, pause, reset. When time ends: a gentle chime using Web Audio (no external files) and the tab title flashes - A session log under the timer: each finished session adds a line with the time and duration, and a running total for today - Clean dark design, works on a phone screen, no sign-up, nothing leaves the browser Explain in 3 bullets how to use it when you're done. If any feature is impossible in one file, say so instead of faking it.
Then the real skill — iterating in plain language:
- “The chime is too harsh — make it a soft two-note tone, half the volume.”
- “Add a field where I name the task before starting, and show the name in the log.”
- “On phones the buttons wrap badly — stack them full-width under the timer.”
- “It broke: after pausing, resume restarts from full time. Find and fix that bug; explain what caused it in one sentence.”
Spec first, code second
For anything bigger than a toy, spend one round on text only: “Before writing any code, write a one-page spec: what it does, the screens, the data it stores, and the 3 hardest parts. I'll approve or edit it, then you build.” Ten minutes of spec saves an hour of “that's not what I meant.”
Make it explain and document as it goes
End build prompts with: “Comment the code like you're teaching a beginner, and keep a short README at the top describing how each part works.” Future-you — and any human developer you later hand it to — will be grateful. Undocumented AI code becomes a black box you own.
AI-built apps are brilliant until they're not: logins, payments, storing other people's data, security, and anything where a bug costs money or privacy. Rule of thumb: if failure would hurt someone who isn't you, a human professional reviews it before real users touch it. Prototype freely; deploy carefully — which is the next chapter.
App Deployment
“It works on my screen” is not deployed. Deployment means your thing lives at a real address, keeps working when you close the tab, and doesn't fall over or leak when a stranger uses it.
The ladder of shipping
| Level | What it means | Good for |
|---|---|---|
| 1 · Share the artefact | Many AI tools give your creation a shareable link directly (Canvas, Claude Artifacts, AI Studio apps). Zero setup. | Showing a prototype today |
| 2 · Static hosting | One-file / static sites dropped onto hosts like Netlify, Vercel, GitHub Pages, or Cloudflare Pages — often a literal drag-and-drop. | Tools, portfolios, guides like this one |
| 3 · App platforms | Replit, Render, Railway and similar run real backends and databases from your project, handling servers for you. | Apps with accounts, saved data, APIs |
| 4 · Production, properly | Your own domain, monitoring, backups, analytics, error alerts. This is where the professional review from Chapter 7 happens. | Anything people depend on or pay for |
The pre-launch checklist (ask your AI to walk you through it)
- Break it yourself first: “List 10 ways a careless or malicious user could break this app or see data they shouldn't. Then fix the top 3.”
- Secrets out of the code: API keys and passwords never live in the app's visible code. Ask: “Where do secrets belong in this setup, and is anything sensitive exposed right now?”
- Test on a phone, on bad wifi, as a stranger: logged out, first visit, no instructions from you standing next to them.
- A way back: keep the last working version. “How do I roll back if this update breaks it?” is a question to ask before updating.
- Know your costs: free tiers end. Ask the AI to map what your app uses and what each piece charges when real traffic arrives.
I built this app with you and want to put it on the internet for real users. Interrogate the project like a cautious engineer: 1. What type of hosting does this actually need, and why? Give me the simplest option that works 2. Step-by-step deployment instructions for that option, assuming I'm a beginner 3. Everything in this project that would be visible to the public — flag anything that shouldn't be 4. What breaks first if 100 people use it at once? 5. A rollback plan if the launch goes wrong Ask me any question you need answered before I start.
Deployment is a spectrum, not a ceremony. A shared link that helps your team today beats a “proper launch” that never happens. Ship at the level the audience needs — and know which level you're at.
The Prompt Template Library
Twelve templates covering the jobs people actually do with AI. Bracketed parts [like this] are yours to fill. Every card copies with one click — paste into any AI and go.
I want to [task]. Before you start, ask me up to 4 questions whose answers would most change your approach. Then wait for my answers before producing anything.Here is my draft: [paste]
1. Summarise in one sentence what you think I'm trying to say
2. List anything ambiguous, contradictory, or missing
3. Ask me about those points — then, and only then, rewrite it in my voice, ~30% shorter, facts unchangedResearch [topic / question] for me.
- Give me the landscape first: main positions, key terms, who's credible here
- Every factual claim needs a source link I can open; prefer primary and recent sources
- Separate "well established" from "contested" from "speculation"
- End with: what would change the conclusion, and the 3 sources I should read myselfTeach me [topic] like a Socratic tutor. First ask what I already know. Then teach by asking me one question at a time, correcting me kindly but honestly, building on my answers. Never lecture for more than 3 sentences. If I get stuck twice on the same point, explain it with a concrete everyday analogy, then quiz me again.I'm deciding whether to [decision]. Context: [what matters to you, constraints, timeline]. 1. Steelman the case FOR and the case AGAINST, equally strongly 2. What am I probably not considering? 3. It's 2 years later and it went badly — write the most plausible story of why 4. What cheap experiment could I run in the next 2 weeks to learn the most?
Help me write a hard message. Situation: [what happened]. Recipient: [who, and your relationship]. I need them to [outcome] without damaging the relationship. Give me 3 versions: direct, diplomatic, and warm-but-firm. Facts exactly as I stated them — invent nothing. Each under [120] words.
Here are raw notes / a transcript from a meeting: [paste]
Produce: (1) a 5-line summary, (2) decisions made — only ones actually stated, (3) action items as a table: task / owner / deadline, using "UNSPECIFIED" rather than guessing, (4) open questions nobody resolved, (5) anything two people seemed to disagree on without saying so.I'm attaching/pasting data about [what it is]. Step 1: audit it — duplicates, gaps, inconsistent formats, outliers. Don't analyse yet. Step 2, after I confirm: show your working for every number, find the 3 most interesting patterns, propose a plausible explanation AND a boring explanation for each, and tell me what extra data would settle it.Goal: [goal]. Constraints: [budget / time / who it's for]. Phase 1: 15 ideas, forced across 3 categories I wouldn't naturally mix — cheap, experiential, and slightly absurd-but-doable. One line each. Then I'll pick favourites and you'll score them against my criteria and argue against your own recommendation.
Create an image: [subject and scene, specifically]. Style: [photorealistic / watercolour / flat illustration…]. Composition: [viewpoint, what's in foreground vs background]. Light & mood: [time of day, feeling]. Details that must appear: […]. Must NOT appear: text, logos, watermarks, […]. Aspect ratio: [16:9 / square / portrait].
First write a one-page spec (no code): what this tool does, its screens/sections, what data it stores and where, and the 3 hardest parts. Tool: [describe the tool and who it's for]. I'll approve the spec, then you build it as a single HTML file, commented for a beginner, with a short how-to at the end. If part of my request is impossible in one file, say so in the spec instead of faking it.Rewrite this so it sounds like a specific human wrote it: [paste text]
Rules: remove AI clichés ("delve", "in today's fast-paced world", "furthermore"), vary sentence length dramatically, replace one abstraction with a concrete detail, cut any sentence that could appear in any article on any topic. Keep my meaning and my facts. Show me what you cut and why, in 3 bullets after the rewrite.The Hack Sheet
Twenty moves, collected from every chapter, that separate people who “use AI” from people who get startling results out of it. Skim it monthly; steal one at a time.
Append to any big request. Converts guessing into briefing.
Paste samples of what “good” looks like instead of describing it.
The antidote to AI agreeableness. Works on plans, decisions, drafts.
Range first. Your taste does the selecting — that's the fun part anyway.
“You assumed X; it's Y” fixes ten downstream errors at once.
Child, student, practitioner — then ask what the expert version hides.
“It failed. Write the story of why.” Then rebuild so the story can't happen.
Describe your writing style once as rules; paste the card into every writing task.
Agree the outline / spec / narrative first. Words are cheap to change; structure isn't.
“A source link for every claim, exact quotes for every number.” Then open the important ones.
Shown calculations plus a second-method verification catches most numeric errors.
Duplicates, gaps, outliers first. Pretty charts of dirty data are still wrong.
State what to exclude — in prompts (“no jargon”) and images (“no text, no logos”). Exclusions are specifications.
“For a sceptical CFO” changes vocabulary, evidence, and length more than any tone adjective.
“Under 90 words” forces prioritisation. Unconstrained AI defaults to thorough and bloated.
Long chats accumulate confusion and early mistakes. Start new for a new task; paste a summary if context must carry over.
Socratic mode for learning, expert-witness mode for decisions: one question at a time, honest corrections.
One good source → summary, posts, headlines, script. Consistency comes free; new facts stay out.
Ask about any plan, draft, or analysis. Models volunteer what's present; omissions need an invitation.
When a prompt works beautifully, save it verbatim (like this library). Prompting skill compounds — but only if you keep the assets.
Responsible Use & AI's Limits
A guide that only taught the power moves would be a sales brochure. Here's the other half: where these tools fail, what they can cost you, and the habits that keep you in charge.
It will sound certain while being wrong
Hallucination isn't a bug that one update will fix — it's what happens when a system built to produce plausible text meets a gap in its knowledge. It invents citations, dates, quotes, and “facts” in the same confident voice as true ones. Your rule: the more a claim matters, the more it gets verified against a source you opened yourself. Names, numbers, dates, and anything medical, legal, or financial get checked, every time.
It's a mirror — including of bias
Models learned from human text, with human blind spots baked in: stereotypes in examples, majority viewpoints presented as consensus, your own framing echoed back at you (ask a leading question, get a led answer). Counter-moves from this guide that double as bias defence: ask for the strongest opposing view, ask what a different audience would object to, and never let it be the only voice in a decision about people.
Privacy: what you paste may not stay yours
Treat consumer AI chats like a semi-public notebook. Don't paste passwords, financial account details, other people's private messages, confidential work documents, or anything you'd be harmed by seeing in a breach or a training set. Redact names and numbers when the substance doesn't need them — the analysis works just as well on “Employee A.” Check your tool's settings for training-data opt-outs, and your employer's policy before using AI on work material.
Your name goes on the output
If AI helped write it, research it, or build it, you are still the author of record — its errors are your errors when you ship them. Disclose AI assistance where your workplace, school, or publisher requires it (many do, and rules differ). And respect others' work: AI output can echo copyrighted text and styles; passing off generated work as wholly hand-made, or feeding in material you have no right to use, are your choices to answer for, not the model's.
When NOT to reach for AI
- Emergencies and crises — call the actual service, hotline, or human. Speed and accountability matter more than fluency.
- As the final word on health, legal, or money decisions — use it to prepare questions for a professional, never as the professional.
- When the thinking IS the point — learning, grieving, apologising, deciding who you want to be. Outsourcing the process destroys the product.
- Judging people — hiring, grading, accusing. Fluent pattern-matching is not evidence about a person.
- Anything you'd need to be true — if being wrong is expensive, AI is a drafting partner, never the verifier.
Read outputs like a good editor reads a talented intern's work: fast, fluent, eager to please — and checked before it leaves the building. Trust is earned per-claim, not granted per-tool. Keep that stance and you can use AI aggressively and safely; lose it, and the hacks in this guide mostly make you wrong faster.
You did it. Here's your certificate.
You read the whole guide — the formula, the eight skills, the hacks, the limits. That is, by any honest measure, the entire content of the paid certificate, minus the paywall and plus a few things it doesn't teach. So here's the only part that was ever really missing: the ridiculous, gold-foil, hang-it-above-your-desk proof.
Certificate of Completion
awarded for mastery of
The Art of Being Incredibly Awesome
- Skill level
- Legendary
- GPA
- 4.0 in Fun
- Serial No.
- 007
- Date
- October 2026
Your actual credential
Forget the gold foil — the real proof is portable: open any AI, paste Template #1 from the library, and watch what happens. Skill you can demonstrate beats paper you can frame, every time. (Frame the paper anyway. It's magnificent.)