A system you keep

Start from the hours, not from a number of posts. Write down the hours you can protect in an ordinary week, not a good one. Estimate how long one piece at the depth that earns a result takes you: a teardown, a worked example, a considered answer. Divide. The result is your cadence, and it is usually smaller than the advice.

Then cut it by a third for the weeks that go wrong. A cadence you keep through a bad month beats a better one you abandon, because session 4 found no sign that a burst grows the next block, and the per-piece value of what you publish does not rise with the count.

The last chapter title of Austin Kleon's Show Your Work! is "Stick around". It is the honest summary of what the bench can and cannot say. The bench cannot show that any cadence grows an audience; it can show that the creators who were still answering at the end are the ones every success chart is drawn from (51 of 302). Staying is the precondition for being in that set at all. A small cadence is how people stay.

The disagreement from session 4 still stands: Buffer's LinkedIn analysis found more posts went with more reach per post on a feed platform. If your outpost is a feed, test it on your own numbers for a quarter rather than adopting either finding.

Questions you get asked. If a colleague asked it, someone in your audience is wondering it. Work you finished. Every finished piece is a teardown candidate once anything confidential is removed (session 3). Mistakes you fixed. The most useful public writing in most technical fields is a mistake explained by the person who made it. Reading you disagreed with. A disagreement with a source, stated fairly, is a point of view, which is what a positioning line needs behind it.

Capture is one line, at the moment, in one place. A bank spread across four apps is four banks, and the empty week is the week you cannot find any of them.

The triage exists to protect the line from session 2. An idea that is interesting but does not serve the audience in your line is a good idea for somebody else. An idea you cannot back with a proof is a claim, which session 3 moved you away from. Neither is deleted; both go back with a note. Some of them become right when your work catches up with them.

Turning one idea into several formats is a production skill, taught in AI Content ↗. Here the job is only choosing which idea is next.

Same two searches, same private signed-out window, same ten slots, same four letters, same two counts. Change nothing about the method, or the comparison measures your method instead of your reputation. Then answer three questions in writing: which new result arrived and from which piece, which old result is still wrong, and what the next quarter's one added result will be.

Search results move slowly and for reasons you do not control. A quarter with no change is a data point, not a verdict; two quarters with no change on the second count is a sign that the work is not being published where a search can see it, which sends you back to the platform map from session 5.

Every review should remove something. An outpost nobody in your audience reads, a format that takes twice the hours of the others, a recurring piece that never produced a reply. Stopping is how a small system stays small enough to keep. Write the stopped item in the review with the reason, so a later quarter does not restart it out of guilt.

For managers coaching a team: run the review together, once a quarter, as a short working session. Each person brings their two counts and one result they are proud of. Nobody is compared with anybody else; session 4 showed what cross-person comparisons do to a reading of cadence.