The honest version of "company culture" isn't the deck on the careers page. It's the thousand small operating calls a founder makes in the first three years that nobody writes down: who got hired and why, who didn't, which incidents got handled with grace and which didn't, what tradeoffs the team took versus the ones they backed away from. By year four, most of that record is in nobody's head. The new VP-of-something asks "how do we usually handle this?" and the answer is a guess.
A working notes setup doesn't manufacture culture. It captures the operating record so the culture you're actually building is legible later — to you, to a new hire, to the next leader, to the version of yourself that needs to remember why a decision was made the way it was. The same instinct underwrites how to document your startup's pivot history — the company's lineage and the company's culture both decay if nobody keeps the record.
A vault that mirrors how culture actually accumulates
The shape that holds up is one top-level page for each pillar — principles, hiring, how we make decisions, founding stories, incidents and how we handled them — with sub-pages for the granular detail underneath. Capy supports unlimited page nesting, so the principles page can fan out into one page per principle, the hiring page into one page per role, and so on, without you having to commit to a final structure on day one.
Plain markdown matters here because the agent can read across the vault when somebody asks the cross-cutting questions. "Trace every time we've referenced the 'high-trust low-process' principle in an actual decision over the last year" is a question that's only answerable when the agent can read every page at once.
Principles, with the actual decisions that anchor them
Most company values pages drift toward platitude because the values aren't anchored to anything. "We value transparency" doesn't survive contact with the next hard call.
The fix is to anchor each principle to specific decisions where it was applied, in the actual language of the moment. A working setup is one page per principle with two sections: the principle itself in plain prose, and a running list of decisions where it shaped the call. After each significant decision, ask Capy to read the decision page and propose which principle (if any) it should be filed under, and a one-paragraph entry for the principle's running list.
Six months later, when somebody new asks what "high trust, low process" actually means at this company, the answer isn't an aphorism. It's seven specific decisions where the principle did real work — and seven where you considered it and went the other way, with the reason recorded.
Hiring rationale that survives the rolloff
The most expensive cultural artifact in a young company is hiring decisions, and the worst-kept. The original interview rubrics, the debrief notes, the reasons you said yes to candidate A and no to candidate B — that record is usually in the founder's email and decays the day they hand the function off.
A working setup is a hires inline database on the hiring page, with rows for date, role, hire, the rubric criteria they hit hardest, the rubric criteria you flexed on, and the reasoning. The database lives directly in the markdown page via the :::database::: directive, so the rationale lives alongside the prose context — the principles, the stage of the company, the team shape at the time. After each hire, ask Capy to read the debrief notes and propose an entry; you confirm or edit before it lands. The same per-hire retrospective pattern is the spine of how to use AI notes for hiring.
When the next hiring manager picks up the function and asks "why did we hire for this profile last time," the answer isn't a vague "we wanted someone who'd fit." It's a row with the actual rubric and the actual tradeoff.
Founding stories that stop being myth
Every company has a small set of founding stories that get told to the team — how the first customer signed, the day the team almost broke, the call that pivoted the product. These stories shape culture because they encode what's celebrated and what's tolerated. They also drift, because they live in the founder's verbal retelling and get rounder with each telling.
Park them on a founding stories page. Where there's source material — an email thread, a call recording, a slide deck from the time — drop it on the relevant story sub-page. Capy can transcribe the call recording with speaker diarization (labels like Speaker 1: … so it's clear who said what) so the story is grounded in the actual conversation rather than the founder's reconstruction. PDFs auto-convert to markdown via docstrange, which means the original investor email or the early customer's first response is searchable text the agent can quote back.
The story stays a story; it just stops being myth. The new hire reads the original material instead of hearing the rounded version.
A decision rituals page
How decisions get made is one of the strongest cultural signals in a company, and it's almost never written down. Whether the founder asks for input before a decision or after. Whether dissent is invited or merely tolerated. Whether reversibility is treated as cheap or expensive. The team learns the answer through pattern-matching across dozens of decisions; new hires walk in blind.
A decision rituals page captures the actual ritual: who's consulted before a directional product call, how a hire/fire decision is made, how a pricing change is rolled out, what gets a board pre-read versus a post-fact update. Each ritual links to a few illustrative past decisions where it was applied. Ask Capy to read across recent decision pages and surface where the ritual was followed and where it was bypassed — the deviations are usually more informative than the conformity.
This is the page that turns "you'll figure it out" into "here's the actual shape." Onboarding a new senior hire takes a fraction of the time when this page exists.
Incidents and how we handled them
Every company has a handful of incidents that revealed what it actually was — a security event, a key departure, a customer crisis, a near-miss with a regulator. These moments shape culture more than any all-hands speech. They also tend to get sanitized in the official record, which strips the lessons.
A how we handled it page per incident captures the timeline, the decisions made, the things that went well, the things that went poorly, and the changes made afterward. Drop the relevant Slack threads, post-mortems, and call recordings on the page. Ask Capy to draft a one-page narrative grounded in the actual material — what happened in what order, who decided what, what we'd do differently. You edit it for the version of the story you'd want a new hire to read in their first month.
The incidents stay legible. The lessons survive the people who lived them.
A culture handbook the agent helps you keep current
The culture handbook problem is that it gets written once, three years too early, and never updated. By year five it's wrong about everything that matters and right about the trivia.
Inverting this: don't write the handbook directly. Write the underlying pages — principles, hiring rationale, decision rituals, founding stories, incident narratives — and ask Capy to draft the handbook from them. Once a quarter, ask the agent to read across the vault and produce a refreshed handbook draft, flagging where the current handbook diverges from what's actually in the operating record. You edit, ship, and move on.
The handbook stays current because its source material does. The cost of an update is closer to a coffee than a Saturday.
What this isn't
Capy isn't HR software, isn't a culture-as-a-service product, and isn't a substitute for actually building the culture in the room every day. The vault holds the record of the culture you're building — the decisions, the rationale, the stories, the incidents. The culture itself still gets built by who you hire, who you fire, who you promote, and what you tolerate.
It's also single-user by design. The founder (or whoever owns the cultural record) keeps the vault. Outputs ship to the team as the handbook, the principles doc, the onboarding pages — but the working notes underneath stay in one person's hands. That's the right shape for the messy, in-progress operating record.
A small first test
The cheapest way to see whether this fits is to pick one principle you've been telling the team about — say, "we make small bets fast" — and try to write the underlying page. Pull together five decisions where the principle was actually applied, drop the relevant context onto sub-pages, and ask Capy to read across them and tell you what the principle has actually meant in practice. If the synthesis is sharper than what you'd say off the cuff at the next all-hands, you've got a sense of what the rest of the culture record could look like. (For the offsite-to-incident edge of culture work, AI notes for company retreats and offsites covers the recording-to-recap loop.)
Try Docapybara free. Load one principle and the decisions behind it, and see what the agent does with the operating record.