In short: Somewhere between the third and the tenth hire, the company gains people who were not there when anything was decided. The enterprise brain's job becomes answering "why do we do it this way?" without a founder in the room.
Key points
- If the number of questions founders answer each week grows with every hire, founders have become the company's memory.
- Rewrite ledger entries so a stranger can follow them: context first, no private shorthand, links to evidence.
- Write 'why we do it this way' pages for the five to ten practices new people question most.
- Ten people is the first point where an AI assistant over the record is clearly worth it.
- Topic wikis go stale because nobody owns them; time-ordered decision records are superseded, not updated.
The first employees join a company whose habits, choices and quirks all have reasons, and almost none of the reasons are written down. They ask. The founders answer. That works for a while, and then it becomes the founders' main job without anyone noticing.
The symptom is easy to spot. Count how many questions in a week end with a founder answering them. If that number grows with every hire, the founders have become the organisation's memory, and memory does not scale by working longer hours.
What changes
The founders' ledger stays, but it must now be readable by people who were not there. That means writing entries so a stranger can follow them: context in the first line, no private shorthand, links to the evidence.
Three additions earn their place.
"Why we do it this way" pages. Short explanations of the five to ten practices new people question most: pricing, how leads are qualified, why a particular tool, why a market was deprioritised. Each links to the ledger entries behind it.
First definitions. The handful of terms that already cause confusion. What counts as a qualified lead. What "launched" means. When a customer is "active". Write them down once, with an owner.
Onboarding through the record. Every new hire spends part of their first week reading the ledger. It is the fastest way to absorb how the company thinks, and their questions reveal which entries are unclear.
AI at this stage
This is the first point where an assistant over the record is clearly worth it. A new hire asks "why did we stop selling to agencies?" and gets the ledger entry, the date and the reason. The question no longer interrupts a founder. If the record cannot answer, the question is logged, and the founder writes the missing entry once instead of answering it five times.
The trap
The trap at ten people is a wiki. Someone sets up a large knowledge base, populates it enthusiastically for two weeks, and then it decays. Wikis organised by topic go stale because nobody owns keeping them current. A decision record organised by time does not go stale in the same way, because old entries are meant to be old; they are superseded, not updated.
In practice
- Count questions answered by founders each week; aim for that number to fall as headcount rises.
- Rewrite ledger entries so someone who was not there can follow them.
- Write "why we do it this way" pages for the five most-questioned practices.
- Make reading the ledger part of every new hire's first week.
What would prove this wrong. Ten-person companies where founders answer every contextual question without strain or slowdown. Then the record is premature at this size.
Common questions
How do startups stop founders being the bottleneck for every question?
Make the founders' ledger readable by people who were not there, write 'why we do it this way' pages for the most-questioned practices, write down first definitions with an owner, and onboard new hires through the record.
Should a 10-person startup set up a wiki?
The article calls the wiki the trap at ten people. Topic wikis decay because nobody owns keeping them current, while a decision record organised by time does not go stale the same way.
When is an AI assistant over company knowledge worth it?
At around ten people. A new hire can ask why something was decided and get the ledger entry, date and reason, and unanswered questions are logged so the founder writes the missing entry once.
What should new hires read in their first week?
The decision ledger. It is the fastest way to absorb how the company thinks, and their questions reveal which entries are unclear.