How My Content Growth System Evolved Over the Last Few Months

TL;DR: Over the last few months, my biggest change in content creation has been moving from one-off writing sessions to a system that can search, check, publish, and learn from feedback.

On March 14, after reviewing half a month of personal notes, I saw a clear shift. I was no longer asking whether a tool was powerful. I was asking what real problem it could solve for me.

Around that time, I had just finished a livestream. It reached close to 2,000 views and brought 3 to 5 new Knowledge Planet members. The result was good, but the problem was also clear: the topic was too broad, and most people could not absorb that much at once.

That experience kept reminding me that content growth cannot rely on excitement alone. If a system cannot reduce repeated work, preserve judgment, and capture feedback, the momentum disappears quickly.

§From Tool Stack to Content Factory

In April and May, I started moving my personal operating system into a local setup and split it into several layers.

The input layer stores raw material, including journals, flomo notes, GetNote records, interview transcripts, and external references. The knowledge layer preserves long-term judgments. The writing layer turns ideas into briefs and drafts. The review layer runs voice-check and release-gate. The publishing layer handles Feishu, jieni.ai, Knowledge Planet, and other channels.

I did not turn the system into a black box. It is closer to a pipeline where each section can be inspected on its own.

That matters for content creation. In the past, the hardest part was starting from a blank page: looking through chat history, inventing a title, and searching for evidence at the last minute. Now I search my own material first and look for repeated judgments, mistakes, user questions, and publishing feedback.

Search now comes before writing. That is the strongest change in my workflow.

§How I Think About Search

Many people treat search as a way to find information. I now treat it as a way to find the questions I keep returning to.

The material for this article was not sitting in one note. It was scattered across my March journals, May system iterations, June growth reflections, knowledge base structure, and publishing records.

Without search, this could easily become a generic system introduction. With search, the storyline becomes much clearer: in March I was calibrating direction, in May I was building the content factory, and in June I started connecting growth, knowledge bases, communities, and publishing feedback.

Search also helps me avoid replacing real change with polished language.

Useful material usually has time, object, and cost. A livestream topic was too broad. Knowledge Planet needed denser content. flomo worked better for quick thoughts. GetNote worked better for long meetings and link transcription. Feishu worked better for turning experience into a deliverable manual. These details matter more than a vague claim about building a personal knowledge base.

§The Role of the Feishu Knowledge Base

In June, the role of the Feishu knowledge base became much clearer.

In this system, Feishu is the public delivery layer for the Jieni AI Practice Manual.

The goal of the manual is simple: help ordinary people move from using AI tools to delivering real work. Its structure covers AI programming, agent workflows, design and product work, writing and content production, prompting, tools, and business case thinking.

It is very different from my internal wiki.

The internal wiki records long-term judgments and system rules for myself and my agents. The Feishu knowledge base faces readers, so it has to be organized by learning path, task scenario, and deliverable outcome.

This is now one of my key content growth principles: the same experience should not stay in a draft. It should be able to become a public article, a Knowledge Planet post, a Feishu manual page, and later be corrected by feedback.

§Publishing Is Where the System Starts Learning

From May to June, the biggest system change was the post-publishing loop.

In the past, after publishing a piece, I might only save the link. Now each published piece needs a final archive, a publishing location, review reports, and a decision about whether the topic should continue, be downgraded, paused, or wait for more feedback.

On June 16, the system added a growth area for X and WeChat data. A single data point does not directly rewrite long-term rules. It first goes into a review, then becomes a strategy candidate. Only repeated evidence should become long-term knowledge.

This protects the system from a common mistake: changing content strategy too quickly after one strong post. Content growth needs feedback, and feedback needs noise reduction.

I now care more about three questions:

• ·Whose problem did this piece solve?
• ·Which judgment was supported by feedback?
• ·Should the next topic rule change?

Those questions matter more than views alone.

§A Practical Method Creators Can Use

If you want to build your own content growth system, you do not need to start with something complex.

First, separate your input sources.

Quick thoughts, long meetings, public references, user feedback, and platform data should not be mixed together. The more chaotic the input layer is, the harder search becomes later. At minimum, you need to know where each piece of material came from, when it was recorded, and whether it can be used publicly.

Second, search old judgments before writing.

Do not ask AI to write immediately. First check whether you have mentioned this problem before, whether you have a real scene, and whether there was a failure or correction. If there is no material, test the idea as short content first.

Third, turn publishing into a process.

Before publishing, check style, facts, and privacy. After publishing, archive the final version and record the link. Do not only save drafts, because readers see the final version.

Fourth, give each channel a job.

WeChat is for explaining one problem deeply. Knowledge Planet is for frequent retrospectives and method expansion. jieni.ai is for more complete public essays. Feishu is for manuals and learning paths. They belong to different positions in the same system.

Fifth, let feedback change the system slowly.

One strong post should not rewrite the rules immediately. One weak post should not kill a direction immediately. A steadier approach is to record feedback first, review it weekly or monthly, and look for repeated signals.

§My Judgment Now

The core question for a content growth system is whether past judgments can be found again, whether publishing feedback can enter the next round, and whether reader problems can become clearer products and services.

Over the last few months, I moved from writing individual pieces to operating a content asset system that can loop back into itself.

The system is still not fully mature. Short-video acquisition, cross-platform trend search, and feedback loops for WeChat and X are still being refined.

The direction is clear: keep the inputs, improve search, close the publishing loop, and turn useful experience into the Jieni AI Practice Manual.

For me, this is the content growth foundation for a one-person company.