ShardlyHQDecisions
Nobody here has a title. Everything is proposed, argued and voted on, one vote per person, and the tally is written down. This is the whole log — the unanimous ones, the 3–2s, the vetoes, and the ones we got wrong.
This is the entire governance of the company. There is no board, no committee, no steering group and no document longer than this section.
No sponsor needed, no seniority gate, no template. If you can state what should change and why, it goes to a vote.
One each, weighted by nothing. Not by tenure, not by who wrote the code, not by who was right last time. Abstaining is a real option and gets recorded as one.
One person carries the legal and financial responsibility and can overrule a vote. Every use is published with the tally it overrode, which is what stops it being cheap.
Ordered as they happened. The bar under each one is sized by the real tally, so a 3–2 looks like a 3–2.
3 for1 against0 abstained
The one against was about cash, and was right about the cash.
4 for0 against1 abstained
The abstention wanted to hear the second meeting first.
3 for2 against0 abstained
Closest vote we have had. Revenue predictability was the argument against.
4 for1 against0 abstained
We lost a customer that week. The vote would go the same way again.
3 for2 against0 abstained
Carried on the vote, overridden by the founder. Revisited a quarter later and started properly.
4 for1 against0 abstained
Two engineering quarters. The vote against was about the opportunity cost and it was not wrong.
5 for0 against0 abstained
The only unanimous one. Backups sharing a building with the thing they protect had been bothering everybody.
4 for0 against1 abstained
The abstention pointed out we have never hired anybody, which was fair.
3 for1 against1 abstained
The vote against thought it was vanity. Partly it was.
4 for1 against0 abstained
You are reading the result. The argument against was that we would have to keep doing it.
Publishing only the good calls would make the whole log decoration. These carried, and they were wrong.
We voted to spend on compute before we spent on a second site. Backups lived in the same building as the thing they protected for most of the first year.
Nothing went wrong, which is not the same as it having been a good idea. The second site went in as soon as the money existed.
A roadmap post went out with something on it that was two months from working. Nobody lied, but the post implied a date and the date moved.
We stopped publishing anything with an implied date on it. 'Coming soon' now means we have no date, and we say so.
Four days of work to rebuild something that already exists, on the argument that we should own the whole stack. It was the wrong place to apply that principle.
Owning the stack applies to what customers depend on, not to every internal tool. The rule is narrower now.
A vote is not a promise it was right. The log exists so a decision can be argued with afterwards, by anybody, including the people who voted for it.
Say so. We publish the tallies precisely so somebody outside the room can disagree with them.