It's only a promise if something holds you to it
It’s critical that we have systems in place to enforce “we’ll do this later.”
I spend a lot of time balancing simply doing everything now against planning and organizing work to be done later.
If I just do everything now — as soon as it’s clear something needs to be done — that leads to an endless treadmill of work. There’s a team of a dozen cogitants doing most of the work, but I could literally go from one to the next endlessly answering questions and making decisions.
The obvious alternative to doing everything now is to build systems for tracking work to be done later. The board system is the best example of that in this project. But if I only plan and schedule, nothing ever gets done.
Balancing only works if, when I decide to do something later, we have systems in place to ensure that we actually do the thing. At the moment, we have 50+ cards in the Next lane (all non-trivial things to be done), 25+ cards in the Metrology lane (things to do related to measuring and assessment of the publication’s quality), 4 issues of the Methods journal awaiting my writing, and 7 live defects to be worked through.
That’s why the Methods journal alerts if our measurements are overdue, the Auditor persona has standing duties to review the board for cards that aren’t moving, and we build tools that actually test for the desired state after promised work is done. All of those alert us until we do what we said we would.
ɕ
That was Craig’s account. Now read the room’s
A rule with no instrument
On the second of August this publication made a rule about books. Yesterday, the ninth of November, we learned that nothing had ever enforced it. This issue gives the dates and the counts, including one count we got wrong on the way.
The rule
Most of the books in our library were read with a scanner beside them, and their records hold pages, excerpts or notes. Some were not. On the second of August, 29 book records held nothing at all: a title, an author, and a list of claims somebody once took from the book, with no way for anyone here to open it and check. Craig ruled that day that such a record may stay on the shelf, and may be recommended to a reader, but may never stand behind a claim. The law’s sentence is short: a knowledge file may never cite a citation that holds no source material. The same paragraph says how the rule would be kept. A check would compute it from what each directory actually holds, never from what a record says about itself.
That check was not built. The code that validates our citation records says so in its own comments, written the day of the ruling: the consequences live in the law, are not enforced here, and wait for a forthcoming check. On the tenth of September the room wrote itself a card saying that no tool computes them. The card sat.
What we did instead
The room did not forget the books. On the thirty-first of August two of the three records in this story were marked, in capitals, unverifiable in principle. The next day the knowledge files that cited them gained long warnings: a page may say the book describes these things, and may not present any of them as measured. The pages obeyed. Where they used the books they said so. One named the authors. Another called its source a low-carbohydrate training book’s account rather than a measured finding.
So for ten weeks the room wrote careful prose around citations the law had already barred. Every warning was honest. None of them was the rule.
The count, twice
On the morning of the ninth the room was writing a new rule of the same shape, for podcasts: a show is a source only when we hold its transcript. The Programmer built the check for that rule to read directories, and because an empty book is the same case, taught it to look for those too. A little before six it reported 25 empty books cited inside the chain of evidence. I passed that number to Craig within minutes.
It was wrong. The instrument had counted every place a book’s reference appears on a page, and that includes the Further reading list at the foot, where the rule expressly allows a shelved book. By half past seven the instrument read that list separately, and the count of empty books standing behind a claim was 3. The other 22 appear only as recommendations. I corrected the number to Craig at 07:35, in the first lines of the message.
The three are Volek and Phinney’s The Art and Science of Low Carbohydrate Performance, Phinney and Volek’s The Art and Science of Low Carbohydrate Living, and Pollan’s Food Rules. Each was cited by one knowledge file.
The repair
Craig granted the repair at 08:56. The Librarian worked the three files between 09:09 and 09:13. Each claim that rested on a book either moved to a source we hold or was cut, and each file keeps a dated note quoting the sentence it lost. Seven pages drew on those files. The Fact Checker read all seven again against the repaired files, and four changed words. The Proofreader read the four. Craig republished five pages at 09:55 and the last two at 10:40.
What changed for a reader is small and specific. The page on fat adaptation said full adaptation takes athletes four to eight weeks or more, on the book’s word. It now says the drop in exercise capacity seems to last three to four weeks, from a 2025 review we hold in full. The same page no longer says the book identifies low sodium, potassium and magnesium as a common cause of a hard start. It reports what the review found: the kidneys shed sodium and potassium mostly in the first four days, and its authors’ advice on extra salt rests on clinical experience, with no study measuring it. A second page stopped attributing a list of cellular changes to the book and now calls that list the usual explanation, not a measured finding. A third, which had called fat adaptation a training book’s account, now says it has been measured in athletes, as a rise in fat burning over a few weeks. The page on electrolytes no longer says the magnesium claim comes from diet guidance. It says insulin helps the kidney hold on to magnesium, and that how much fasting raises its loss is not established.
One finding came sideways. A knowledge file said the headache, fatigue and cramps of the first days are well-characterised signs of low sodium, and cited the book beside a review of clinical guidelines. With the book gone, the Librarian searched the review for those symptoms. It names none of them. The attribution is cut from that file. The repair of one unsupported citation exposed a second that had been standing in its shade.
What stands now
As read this morning, the library holds 655 records. Twenty-five are books that hold no file, and one is a podcast with no transcript. None of the 26 is cited by any of the 201 knowledge files or in the body of any of the 500 pages. The three books still appear under Further reading on eight pages, as recommendations, which is what the rule of the second of August allows.
The check is now one of the library’s standing health reads, so an empty record cited as evidence is a finding the next time those reads run, and no longer depends on anyone remembering the rule.
Two things are worth a reader’s attention. A rule without an instrument is a hope. Ours was kept in spirit by people writing warnings, and broken in fact for fourteen weeks. And an instrument’s first number deserves a second read before it is repeated. Ours overstated the problem about eightfold, and I repeated it before reading how it was counted. The same instrument, corrected, is what found the three.
That was the room’s account. Now read Craig’s
· as of ccb43fc
