Find the thing that should be easier.
Your AI asks one question at a time and helps you hold one work frustration until it fits on a napkin.
https://distilligent.ai/napkin/agent
TWO SKILLS FOR YOUR AI · FROM THE TALK
How do I get a stack?
Your AI walks one piece of your work through ten steps and hands back a Stack Card and a skill of your own.
https://distilligent.ai/napkin/stack
Paste a link into ChatGPT, Claude, Gemini or whichever AI you use. If it can’t open links, each section below has the text to copy.
AI for Everybody · October 10 · after the talk
SKILL 1 · THE NAPKIN
Bring the thing
that should
be easier.
One work frustration. Ten minutes with the AI you already use. Leave with a thought clear enough to fit on a napkin.
Make my napkinNo account needed. Share only if you choose.
YOUR TEN-MINUTE EXPERIMENTA small thought.
A small thought.
Some room to unfold.
We made a skill for your AI.
Paste one link into ChatGPT, Claude, Gemini — whichever you already use — and for one conversation it learns to think the problem through with you instead of answering at you. What you keep at the end fits on a napkin.Your own AI. Your own words. Ten minutes.
On October 10, the talk took the room’s napkins and worked them on stage. Yours still counts: make it now, and when it has shape, the Stack skill below helps you work out what a fix would need.
— Izza Masud, your speaker
Paste one link.
If your AI opens links: copy this, paste it into your AI chat, press enter.
https://distilligent.ai/napkin
For AI readers: the plain-text instructions STEP 2 · TALK FOR TEN MINUTES
TALKit asks, you answer
NAPKINyour thought, in your words
SHAREonly if you choose
What’s one part of your work that should be easier than it is?
honestly? getting a straight picture of where every engagement stands without chasing five people…
When you say “chasing” — what are you actually waiting for each time?
huh. now that you ask — it’s usually the same two things…
One question at a time, always on yourthought. It won’t pitch software or rush to fix — the fastest way to a bad tool is solving the wrong problem. Somewhere in the middle, the vague annoyance turns into one clear sentence. That sentence is your napkin.
The goal is not a faster answer. It is letting the thought become something before it gets solved.
WHAT YOU KEEP
Three lines, dated, in your own words.
The conversation stays where it happened — we never see the transcript. Your AI hands you back MY NAPKIN: the thing that should be easier, what makes it hard, and what’s still unresolved. Save that block.
Keep confidential work out of it.No client names, tax records, account numbers, passwords, or confidential firm information.Look at the example — no client, no numbers, still real. Describe the situation without sharing sensitive details.
MY NAPKIN
Started · 8:14 PM
The thing that should be easier
I keep rebuilding the same picture of what is still missing before I can move the work forward.
What seems to make it hard
The information exists, but it is spread across messages, files and earlier decisions.
Still unresolved
Which parts require me, and which parts could safely happen without me?
an example of the three-line formatSTEP 3 · DECIDE WHAT TO SHARE · 30 SECONDS
Your napkin can shape what comes next.
The talk’s opening slides were built from what the room brought, one pattern each. Share yours and it joins those patterns. Or keep it; that’s a real option too.
AFTER THE TALK · HOW DO I GET A STACK?
Let the work choose the stack.
The question from the room was simple: how do I get one? Start with the takeaways, then walk one piece of your own work through ten steps. You leave with a Stack Card, a map of where the risk lives, and a skill your AI can follow.
The cost of finding out has changed.
The napkin was always cheap. The first honest test was not. Now it can be.
AI is stochastic.
Good at messy inputs, summaries, patterns, drafts and options. Never leave exact totals, posting money, policy, irreversible actions or client-impacting decisions unguarded.
Ask “accurate at what?”
Models reason far better than they recall facts. Reasoning goes to the model, lookups go to the source, exactness goes to a deterministic check.
One workflow. Three zones.
A safe system separates judgment (AI), exactness (rules and math) and authority (a person). For every piece of work, ask which zone it is in.
Build a stack, not a prompt.
Sources › Context › AI › Checks › Gate › Action. The model is one part. The harness around it is where work becomes safe and repeatable.
Let the work choose the stack.
Not “which AI should I buy?” but “what machinery does this work need?” A stack is a map of where the risk lives.
Build it Wagile.
Hold the whole idea, build it in short cycles, check along the way: plan, build, check, adjust.
Four words, no fog machine.
Model: the thinker. Harness: the rules, tools, context and controls around it. Skill: a repeatable way of doing work. MCP: a tool or data socket the AI can use, with permission.
Make AI a co-thinker.
Not an oracle. Make it ask, challenge, show its sources and name what it does not know. Inspect it better; don’t trust it more.
Before it bites.
Ask for sources. Try it on realistic sample files. Count the cleanup time. Keep sensitive data in approved systems.
People hold the trust.
AI can prepare the room. You hold judgment, accountability and final authority. Use it to be more present with clients, not less.
Can I verify this?
Your client’s AI may be the next evaluator. Credible work shows its evidence, sources, assumptions and who decided.
Sources›Context›AI›Checks›Gate›Action
Give your AI the stack skill.
Paste this link into the chat where you made your napkin. Your AI walks you through the ten steps one question at a time: your workflow, its zones, sources, model, MCPs, skills, checks, sign-off and action. It hands back your Stack Card and a skill you can keep.
https://distilligent.ai/napkin/stack
Or walk the steps yourself.
The same ten steps, on this page. The card builds as you go. Nothing you type here is sent anywhere; we only count visits and button clicks. Copy what you want to keep.
That’s everything. No account, nothing to install, nothing sold. A working conference experiment from Distilligent / MindSpace, first run at AI for Everybody on October 10. Take one chance. Make it cheap. Make it honest.