Before anything else today, look at this chart.
Each dot is millions of people
2,500 dots · color = most advanced AI interaction
based on estimates shared by Alex Finn, Feb 2026
Grey: never touched AI. Green: free chatbot users. Orange: the few who pay for it.
Red: the tiny group running AI on their own files, with their own system.
Four days ago, you were green.
Congratulations: four emails in, and you're standing at the door of the smallest group on this chart. One foot is already in the red.
(If you've ever felt behind on AI — stop. Look how far the right information, made for you, just took you. You're already ahead of most of your colleagues.)
You can now run a real piece of your job with a single word.
Go ahead — run your skill right now. Your words, press enter. You earned it.
(Notice something: this is a brand-new session. Yesterday's chat is gone. Bruno read his memory file, found your skill on his own, and ran it anyway. That's the system proving itself.)
So today, let's see what you can do that the others can't yet.
The Round Table Framework
I imagine AI as a round table.
Albert Einstein is sitting there. Elon Musk. Miles Davis. Mike Tyson.
And any world-class expert, in any discipline you can name.
That's Bruno. Bruno is any of them, whenever you need one.
Now, here's the thing about that table.
It's the same table for everyone. Same experts, same seats.
What changes is YOU.
Most people ask generic questions.
Would you really ask Einstein how the weather looks?
Or they ask the wrong expert.
Would you ask Elon Musk for marriage advice?
(I love Elon... but I would not.)
Not you. Not anymore.
Your Bruno knows you, your job, your standard — and he already runs a real task on your behalf.
Which means the experts at your table can finally advise you like the professionals they are.
Here's how we put them to work. Five steps:
- Ask the table. Ideas, not changes.
- Pick like a manager. One or two. Judgment stays yours.
- Demand a plan. Nobody touches a working process without one.
- Compare the versions. Old one next to new one.
- Make it permanent. The upgrade goes into the skill.
Step 1. Ask the table.
With the fresh output in front of you, ask the one question managers almost never think to ask:
"What could you do to make this better?"
"Look at the finished work in `output`. You know my task, my standard, and who receives this work. Now review it like a table of the best professionals in the field: what could make it more valuable to the person who receives it? Give me your best ideas in plain language: what you'd change, why it's better for my reader, and what you'd need from me. Ideas only. Don't change anything yet."
Now read the list.
Somewhere in it, there's an idea that makes you stop.
Something useful you never thought to ask for. On your real work.
That's the table earning its seat.
Step 2. Pick like a manager.
Don't take everything.
Ideas are now abundant. Judgment is the scarce part, and the judgment is yours.
Pick one or two that genuinely serve your reader.
Ignore the rest without guilt.
Step 3. Demand a plan.
Here's a rule you already live by.
Nobody changes a working process without showing you a plan first.
Your report works. So Bruno doesn't touch it until you've approved his thinking.
And a money reason to love it.
AI work is billed in tokens: every word read and written costs a little.
Work done wrong and redone costs three times the tokens, and three times your patience.
Correcting a plan costs almost nothing.
You already manage people this way. Now you manage Bruno this way.
Switch Plan Mode on, and paste:
"Let's apply the improvements I chose: [WRITE THE ONE OR TWO YOU PICKED]. Propose your plan, step by step, in plain language, and wait for my approval. After I approve, apply them and save the improved version in `output`, next to the current one. Change nothing else."
Read the plan. Correct the thinking if something's off. Approve.
Step 4. Compare the versions.
Open both.
The one you've been sending for years. And the new one.
Side by side.
That difference is what the table is for.
Step 5. Make it permanent.
"The improved version is the new standard. Update the skill file you created yesterday inside .claude/skills: add the improvements we chose and the reasons behind them, so every future run produces this new version. Keep the skill's name and its frontmatter valid. Change nothing else in my folders. Confirm when it's done, and tell me if my run words stay the same."
Done.
Your button now produces the better version. Every time.
Congratulations.
Look at the map: Problem 3 is gone.
Your task runs consistently, at a standard that just got higher than the one you've been sending for years.
And you did it the manager way: ideas from the table, judgment from you.
The new version exists and it's genuinely better.
Improvements came out messy or too big.
Stuck.
DM me on LinkedIn: DAY 4 + your color.
Yellow today usually means the improvement was too ambitious: shrink it, rerun.
I read everything.
Tomorrow, the last day.
No new installs, no new tools.
Just three things: the full list of what you actually built this week, one last gift...
...and where this road goes next.
Bring a coffee.
Filippo