A publishing system creates a new problem: too many possible things to write.
Once the intake works, ideas arrive from everywhere. Readwise highlights, daily digests, saved articles, profiles, papers, conversations, content candidates, old notes, new questions, half-finished drafts. The bottleneck stops being capture and becomes judgment.
That is the healthier bottleneck, but it is still a bottleneck.
I do not want to write everything that is interesting. Interesting is too low a bar. I want to write things that have some combination of personal pull, reusable insight, source support, and fit with the body of work I am building.
The first filter is energy. Do I actually care about this, or do I merely recognize that it could be content? That distinction matters. A lot of ideas are plausible. Plausible ideas are dangerous because they can fill a queue forever. The better signal is irritation, fascination, repeated return, or a sense that the idea explains something I keep seeing.
The second filter is usefulness. Could this help someone think, decide, inspect, or build differently? I do not mean that every piece needs a checklist. Some pieces are more reflective. But the reader should leave with a sharper lens, not just the feeling that they consumed a competent essay.
The third filter is fit. Does this belong in the notebook's world? The recurring themes are fairly clear: operators, company systems, AI workflows, knowledge work, public learning, profiles as lessons, research as practical judgment, business building, and the strange ways tools change behavior. A piece can be good and still belong somewhere else.
The fourth filter is evidence. What is this based on? A personal essay can stand on lived experience and clear self-limitation. A research explainer needs the source. A profile needs receipts. A company deep dive needs a much stronger source base. The evidence bar should match the claim type.
The fifth filter is novelty inside my own work. Have I already written this? If so, is the new version adding a sharper angle, a new example, a different artifact, or a changed view? Repetition is not always bad. Some themes deserve to recur. But accidental repetition is waste.
This is where accepted and rejected idea feedback becomes useful. If every rejected idea disappears into memory, the system will suggest it again later. If every accepted idea stays in a vague queue, the system will not know what actually moved. The feedback loop teaches the editorial assistant what my taste is becoming.
The decision is not binary. There are at least four outcomes.
Draft now. The idea has energy, fit, and enough source basis. It should move.
Save for later. The idea is good but not ripe. Maybe the source base is thin. Maybe it belongs after another post. Maybe it needs a better example.
Watch. The idea is interesting but still unresolved. It needs more signals before becoming a claim.
Reject. The idea is too generic, too buzzy, too derivative, or simply not mine.
The reject category is important. A good content system needs a trash function. Otherwise the backlog becomes a museum of guilt.
The hardest ideas to reject are the ones that sound strategically correct. "AI will change knowledge work" is true and mostly useless as a standalone idea. "Dashboards are giving way to exception-driven operating loops" is more specific. "How I use agents as editors instead of writers" is even more mine. Specificity is a taste filter.
The best writing candidates usually have a sentence I can say plainly. Not a clever title. A plain sentence. "Profiles are more useful as pattern libraries than biographies." "A public notebook is a learning system with a publish button." "Agents are better as editors than ghostwriters." If the plain sentence is not there, the idea may not be ready.
The workflow helps by surfacing candidates, but it should not choose on its own. Recommendation briefs are useful because they create a small menu. They are dangerous if they become a substitute for wanting to write.
The final decision is still human: do I want this in the public body of work under my name?
That question sounds obvious, but it cuts through a lot. It rejects content that merely performs relevance. It protects voice. It keeps the notebook from becoming a publication treadmill.
Writing compounds when the pieces belong together. The question is not "can I publish this?" The question is "will I be glad this exists when someone reads five other pieces around it?"
That is why the best ideas usually feel connected before they feel finished. A profile might connect to a book note. A research paper might explain a workflow problem I have already seen. A small personal belief might become the introduction to a larger operating essay. The system helps by keeping those pieces close enough that the connections show up.
The decision to write is really a decision to give an idea a place in that web. If it does not belong there, it can stay private, wait, or disappear. That restraint is part of the work.
This is part 7 of 8 in Operating a Public Notebook.