Notes
Think beforeconsensus forms.
Working convictions about what changes next — and what should be built because of it. We publish them early, which means we sometimes publish them wrong.
2026
The wedge is not the company
Start narrow enough to enter. Think wide enough to matter.
↗Intelligence needs a return path
An AI product compounds only when using it makes the next decision better.
↗Proof before permission
A working thing can cross a room that an argument cannot.
↗The person the idea needs
Founder–market fit is almost always told backwards.
↗Move fast where you can return
The speed of a decision should match the cost of changing it.
↗2025
Distribution is not a phase
How a company reaches people is part of what the company is.
↗Start at the center
Bring us the idea before it is safe.
The best moment to talk to us is the moment the old answer stops being obviously right — long before the deck exists. Tell us what changed, and why you cannot stop thinking about it.