SHIPPED01 / 03
Codex4Leaders, in alpha.
Not slides about AI. A workspace the AI itself helps you use, on your own messy leadership work, no code. The first alpha seats are taken, so the next step is the join list.
Join the list →Almost Impossible. Almost.
§ 00 / MASTHEAD
A QUIET CATALOG OF HARD PROBLEMS, NUMBERED & FILED.
The strategic edge of AI, for the people who own the outcome.
AI IS A LEADERSHIP PROBLEM. TREAT IT LIKE ONE.
Almost Impossible.Almost.
FOLIO 001 / 048 · SET IN PLAYFAIR, INTER & JETBRAINS MONOSCROLL ↓ READ THE STORY
AI IS A LEADERSHIP PROBLEM. TREAT IT LIKE ONE.
Almost
Impossible.Almost.
§ 00 / MASTHEAD
A QUIET CATALOG OF HARD PROBLEMS, NUMBERED & FILED.
The strategic edge of AI, for the people who own the outcome.
§ 01 / THE STORY
ENTRY A · ORIGIN · REV. 2026.04
‘Almost impossible.’
The compliment I give to problems that earn it.
When someone brings me a problem that sounds unsolvable, I say the same thing every time. My team in aerospace has heard it. My cohort at ESCP has heard it. Anyone who has spent more than thirty minutes with me on a hard problem has heard it. ‘Almost impossible.’
It comes out as a compliment. A signal that the problem is worth taking seriously, that it sits at the edge of what is actually doable. Not every problem gets this. The boring ones do not. So I started writing about the ones that do.
§ 02 / THE LEDGER
RECEIPTS, NOT OPINIONS
BUILT02 / 03
Tools, on the open web.
I build the things I write about. A leader with no engineers can describe a small tool in plain language and watch it go live with a real link. I know because I do it, then I write down exactly how. The proof is that you can use it, not that I described it well.
Steal the tools →RUN03 / 03
Written from inside the work.
Every piece here comes from running real operations, not watching from the sidelines. Aerospace by day, AI agents on the same board memos and budget reviews I am actually accountable for. If a framework did not survive a real meeting, it did not make the page.
Read the dispatches →Every entry is dated, numbered, and real. If it is not shipped, it is not here.
§ 03 / LATEST
THE MOST RECENT DISPATCHES
001
DOCUMENT · CHINA
JUN 27 · 2026 5 MIN READ
Speed Is Not the Strategy
I went to China expecting a lesson in speed. Shanghai and Shenzhen taught me something more useful: build the conditions that make responsible speed possible.
READ THE ESSAY →
001
DOCUMENT
5 MIN READ
DOCUMENT · CHINA · JUN 27 · 2026
Speed Is Not the Strategy
I went to China expecting a lesson in speed. Shanghai and Shenzhen taught me something more useful: build the conditions that make responsible speed possible.
READ THE ESSAY →
002MAR 6 · 2026The Lego Set for InnovationWhat speaking at the first no-code conference taught me about how we should think about building with AI.DOCUMENT · INNOVATION→
002MAR 6 · 2026
The Lego Set for Innovation
What speaking at the first no-code conference taught me about how we should think about building with AI.
DOCUMENT · INNOVATION→
003MAR 6 · 2026Wrong Idea, Ages to Build, Start OverThe enterprise innovation failure pattern hasn't changed in decades. The technology has. The problem hasn't.DECODE · INNOVATION→
003MAR 6 · 2026
Wrong Idea, Ages to Build, Start Over
The enterprise innovation failure pattern hasn't changed in decades. The technology has. The problem hasn't.
DECODE · INNOVATION→
004FEB 14 · 20264% of the Brain, 100% of the PerformanceA model that fires 4% of its brain and matches the best in the world for $1,892 a year. Intelligence is becoming a utility.DECODE · AI→
004FEB 14 · 2026
4% of the Brain, 100% of the Performance
A model that fires 4% of its brain and matches the best in the world for $1,892 a year. Intelligence is becoming a utility.
DECODE · AI→
005FEB 8 · 2026100 Agents, Zero EngineeringMoonshot trained agent orchestration directly into the model. The application layer companies spent years building may become a model feature.DECODE · AI→
005FEB 8 · 2026
100 Agents, Zero Engineering
Moonshot trained agent orchestration directly into the model. The application layer companies spent years building may become a model feature.
DECODE · AI→
§ 04 / SESSIONS
30 MIN · FREE · ONE SLOT A DAY
Got an almost impossible challenge?
Thirty minutes of honest attention on your hardest problem. No pitch, no agenda, just curiosity.
One slot opens each evening, 18:30 to 19:00 CET. Free, like everything else here.
The best coffee chats become working sessions: one leader, one real AI problem, your hands on the keyboard.
Looking for the tools instead? THE DESK IS OPEN →
§ 05 / THE LETTER
DELIVERED QUIETLY · NO TRACKING · NO SPAM
One letter a month.
What I tried, what broke, what it taught me. Written by hand, once a month. No forwarding of other people’s takes, no pitch, no spam.