पाठशाला Pathshala · संचालन Sanchālan, Operations · Lesson 09 · Start
A writing culture: decisions in documents, not in chat
Chat dissolves and writing solidifies. A decision log, one-page decision memos, a wiki with an owner for every page and a Friday written update let a company remember what it decided and why.
Pathshala, The Founder Library · 11 October 2026 · 7 min read
Ask a company of fifteen why its price is ₹1,499 a month and not ₹999, and the honest answer is often that it was decided in a chat thread eight months ago by two people, one of whom has since left. The thread is somewhere above six hundred other messages. Nobody will scroll to it. So the question gets argued again, and the company pays twice for the same decision.
This lesson sets out four written habits that let a young company remember what it decided and why, how to introduce them in a team that runs on chat and voice notes, and a monthly reading of the record that keeps it honest.
Chat dissolves
Basecamp’s guide to how it communicates states the principle in four words: writing solidifies, chat dissolves. Chat is built for the present tense. It is fast, informal and ordered by time, which makes it excellent for coordinating the next hour and useless as a record of the last year. A decision made in chat is visible only to the people who were in the thread at the moment, findable only by someone who remembers roughly when it happened, and stated in whatever half-sentence closed the argument. The same guide notes that speaking helps only the people in the room, while writing helps everyone, including people who join the company years later.
The costs show up in three places. Decisions are re-argued, because nobody can find the original reasoning. New hires take months to become useful, because how things are done lives in the heads of the people who were there. And when an investor or an acquirer asks why the company did something, the answer is reconstructed from memory and sounds like it. None of this is visible at five people. All of it is visible at twenty-five.
The decision memo
The most useful single format comes from software engineering. In 2011 Michael Nygard proposed the architecture decision record: a short document with a title and a sequence number, a context section describing the forces in tension, a decision stated in full sentences beginning “We will”, a status of proposed or accepted, and consequences, positive and negative alike. The whole document should be one or two pages, written as a conversation with whoever reads it later. Numbers are never reused. And a record is never rewritten when the decision changes: it is marked superseded and linked to the one that replaced it, because knowing what was once decided is itself useful.
Nothing in that format is specific to software. A pricing change, a decision to enter Hyderabad, a choice of payroll vendor or a change to the leave policy fits it exactly. Add two fields for a company: the owner, who makes the decision, and the date. For the decisions that are hard to reverse, the [lesson on reversible and irreversible decisions](/library/decisions-reversible-irreversible-how-fast) sets out how fast to move and what else the memo should hold.
Amazon is the company most associated with memos, and Jeff Bezos’s 2017 letter to shareholders describes six-page narratives read in silence at the start of meetings, and says a great memo probably takes a week or more to write. A startup should not copy the length. It should copy the two habits underneath: the argument is written before the meeting, and the meeting begins by reading it.
The decision log
Most decisions do not deserve a memo. They deserve one line. The decision log is a single page in the wiki with a row per decision: the date, the owner, the decision in one sentence, and a link to wherever it was argued, which may well be a chat thread. It takes thirty seconds per entry and it is the habit that does the most work, because it makes every other written thing findable. A memo gets a log line with its number. A decision taken in a meeting gets a log line before the meeting ends. A decision made in chat gets a log line from whoever made it. A company of fifteen should expect a few entries a week, and within a year the log tends to become the page new hires are sent to first.
The wiki and the written update
The third habit is about how things are done rather than what was decided: the expense policy, the release process, how to onboard a customer, who approves a discount. GitLab calls its version handbook first: write things down first and then communicate that, so the company has a single source of truth, and document any answer that is not yet documented immediately. The behaviour that makes it work is small. When someone asks a question in chat whose answer will be needed again, the answer is written in the wiki and the reply in chat is the link.
Every page has a named owner and a last-reviewed date at the top. Pages without owners rot, and a wiki full of rotten pages teaches people not to trust any of it. When a page is wrong, the person who notices edits it or tells the owner. Duplicates are merged, not tolerated.
The fourth habit replaces a meeting. Every Friday each team lead writes a short update in a fixed shape: what was done this week, what is next, one thing that is blocked and who can unblock it, and one number that moved. It is read before Monday, and the [weekly metrics review](/library/weekly-metrics-review-one-page-one-hour) starts from it rather than from a round of verbal status. The [meetings lesson](/library/meetings-that-earn-their-time) explains why the status meeting it replaces was so expensive.
If it will matter in three months, it does not belong in chat. Write it once, where it can be found, and answer every later question with the link.
Why the writing makes the decision better
The usual case for a writing culture is memory. The stronger case is quality. Paul Graham’s essay Putting Ideas into Words observes that writing about something, even something you know well, usually shows you that you did not know it as well as you thought, and that half the ideas that end up in an essay are ones the writer thought of while writing it. A founder who drafts the memo for a price rise discovers, in the consequences section, the three enterprise contracts that renew at the old price next month. In conversation that gap is covered by confidence. On the page it is visible.
Graham is careful to add that putting ideas into words is no guarantee they are right. The memo does not make the decision correct. It makes the reasoning inspectable, by colleagues now and by the founders later, when the decision has turned out well or badly and the question is whether the reasoning or the luck was at fault.
Introducing it in a company that runs on chat
A writing culture announced in an all-hands lasts about a fortnight. Introduced as four habits, one a week, it lasts. Week one: create the decision log and have the founders add every decision they make, including the ones from chat. Week two: the next decision that is hard to reverse gets a one-page memo, written by its owner, circulated a day before and read in silence at the start of the meeting. Week three: the founders write the five pages new hires ask about most, and start answering questions with links. Week four: the Friday written update replaces one status meeting.
Two cautions for Indian teams. Many people on the team will write in a second or third language, and a writing culture that rewards polish will silence them; reward clarity instead, and say so. And voice notes are not writing. A ninety-second voice note cannot be searched or skimmed by the person who joins next year, so anything that belongs in the log or the wiki is typed, however it was first said.
The monthly reading of the log
On the last Friday of the month the founders spend thirty minutes with the decision log. Count the entries; a month with almost none means decisions are still being made in chat. Open three memos at random and check that each has an owner, a status and consequences that were actually thought through. Mark superseded anything that has been reversed, with a link to what replaced it. Look at the decisions from three months ago and write one line on each that has had time to show its result: right, wrong, or too early to say. Then check five wiki pages for owners and review dates. Once a quarter, read the whole log from the top. It is the closest thing a young company has to a memory, and reading it is how the founders learn which of their instincts to trust.
The habits above are written for a company of eight to forty people. The sources are below; Nygard’s note takes five minutes and its template can be copied as it stands.