The best evidence for whether something works is not how it feels on launch day.
Three months in, here is what actually changed.
The 1:1s Got Better
This is the clearest outcome, and I did not predict it.
A 1:1 is partly a conversation about now — the project, the decision, the tension you need to address. It is also, if you are doing it well, a conversation that is continuous over time: one that picks up threads from previous sessions, that remembers what was said three months ago when the current situation did not yet exist, that notices patterns across many conversations rather than treating each one as a standalone event.
I was not doing that well. I was doing it adequately. I was present, I took notes, I had reasonable recall. But the deeper continuity — the kind that makes people feel genuinely known and thought about — was partial and inconsistent.
After three months, I walk into every 1:1 with a genuine brief. Not a summary of the last meeting. A brief: what this person is working through, what was unresolved in the last conversation, what I noticed that I want to revisit, what context from other sources is relevant to what they are dealing with right now. The system assembles this. I adjust it.
The effect, over time, is that people feel remembered. Not in a formulaic way — there is a difference between a CRM-generated “I see it’s your birthday” and a genuine “I remembered you mentioned this.” Because the memory is real, the conversation is different.
Meeting Preparation Halved
Not because I work faster. Because the context is already there.
Before Cerebro, preparation meant re-reading notes, reconstructing context, and often asking someone what happened in the last session because I could not find the relevant material in time. It was rebuild work — taking existing information and re-assembling it into usable form before the next meeting.
After Cerebro, preparation means adjusting. The context is already assembled. The relevant decisions, the history, the things in tension — they are there. I review, I add what is new, I go.
The time difference is not the point. The quality difference is.
Decisions Better Documented
When I make a decision — to restructure a team, adopt a tool, kill a feature, change a process — I now record not just what I decided but why. The constraints that made other options worse. The assumptions I was making. The thing I was uncertain about and decided to accept anyway.
At the time, this feels like overhead. Three months later, when a similar decision comes up, or when someone asks why things are the way they are, or when I need to evaluate whether an assumption I made six months ago still holds — the record is there.
This is not a new idea. It is rarely practised, because the friction of capturing decisions is high and the payoff is deferred. Cerebro reduced the friction enough that I actually do it.
What It Changed About How I Think
Writing things down to capture them forced me to understand them first. Not always — sometimes I capture a fragment and return to it. But the discipline of structuring a note, of stating what I understood and why, turns out to be a thinking act, not just a recording one. The process of writing clarified the thought. The vault was supposed to be the output. It became part of the input.
I think more clearly about things I have written about. I think in the direction of making things writable. The vault changed the thinking, not just the storage.
What It Costs
None of this is free.
Maintenance takes time. There are weeks when I fall behind and the vault becomes stale — and the AI layer becomes less useful as a result, reasoning over context that is no longer current. Catching up is not difficult, but it requires discipline that does not always coexist with a full schedule.
There are sessions that break down: where context loaded incorrectly, where the model misremembers something, where the output is confidently wrong. These are rare but instructive. Trust requires verification.
The system only works as well as the data in it. Which means it only works as well as I commit to it.
Post 4: what I built wrong, and what I am still getting wrong.
Leave a Reply