After Shipping md2wechat 3.0, I Had to Face Its Ceiling

TL;DR: I shipped md2wechat 3.0 on July 12. Ten days later, I had to face the market ceiling of a standalone WeChat formatting product. I am still building it, but I now see a narrower role for the product and a wider path for the work around it. md2wechat can remain the execution layer for WeChat publishing while I test workshops, consulting, and hands-on system setup as a possible second growth curve. None of those paths has product-market fit yet.

---

I shipped md2wechat 3.0 on July 12.

I had spent two to three weeks planning, building, and debugging the release. The multi-image layout alone took close to a month. I also reviewed the advanced layout modules and removed several that had become overdesigned.

On release day, 3.0 felt like a milestone.

Ten days later, I wrote a much less comfortable note in flomo: the product may have passed its recent peak, and I needed to reassess both the size of the market and the strength of its moat.

Both reactions were honest.

§Version 3.0 Finally Matched the Product in My Head

md2wechat started as a Markdown formatting tool. By 3.0, it could connect title selection, cover images, infographics, multi-image layouts, body formatting, previews, and WeChat draft creation.

My goal has been to build an agent-native publishing engine. Codex, Claude Code, or WorkBuddy should be able to read a draft, identify what is missing, and call md2wechat for the next step. A successful action should return a clear result. A failed action should return enough information for the agent to investigate. When the agent cannot recover safely, it should ask the person using it to decide.

Version 3.0 was the first time that full path felt real.

It also cost more than the release notes show. A multi-image layout can look fine on a desktop and break inside a WeChat mobile preview because of spacing, dimensions, or unsupported styles. I spent nearly a month going back and forth on those details. As the layout library grew, some ideas added more complexity than reading value, so I cut them.

The product became better. The business questions stayed.

§People Still Asked About It, but Fewer Chose to Pay

In late July, I reviewed the conversations and signals I had been seeing.

People were still asking about md2wechat, but conversion felt weak. Free alternatives had become easier to find, and open source projects were appearing quickly. WeChat formatting is already a narrow market. When budgets get tighter, many people will accept a free option that is good enough.

This is a snapshot from my own product, not a market study. I do not have enough long-term data to claim that the pattern will hold.

The part I can say with confidence is narrower.

Adding more layout modules will not solve distribution by itself. The technical moat is limited because other builders can reproduce many of the visible features. The year of product work still matters: user feedback, documentation, search visibility, publishing workflows, and all the edge cases around agents. The commercial scale of a standalone formatting product may still be small.

Admitting that was disappointing.

md2wechat has been the product I invested in most consistently over the past year. While building 3.0, I was still thinking about how to make it the most complete option in its niche. Soon after release, I had to remind myself that past effort was not a reason to keep spending time in the same way.

§A One-Hour Consultation Revealed a Larger Job

The day before the 3.0 release, I spoke for an hour with a friend who runs an industry analysis media business.

He has a small studio, attends industry events, and has strong editorial judgment. His first question was much more tactical: should he use Codex, Claude Code, or WorkBuddy?

Once we walked through his process, tool choice became a small part of the problem.

He needed a content growth system that could keep running over time. His source material needed a local structure. Industry news had to enter topic selection. Existing work had to become readable by agents. Drafts needed review and distribution paths. Publishing feedback needed to return to the next decision.

The content factory, local knowledge base, Feishu manual, and md2wechat work I had done over the previous six months covered much of that path. md2wechat handled the WeChat execution step. Research and editorial judgment came before it. Distribution, metrics, and review came after it.

That conversation changed how I interpreted tool questions. A person may ask which agent to install when the real friction lives across an entire workflow.

§Giving md2wechat a More Precise Role

I am not shutting md2wechat down. I also do not want to prove growth by adding another batch of low-impact features.

I now see it as one component in a broader content system.

The WeChat execution path still needs maintenance. Installation, formatting, images, previews, draft creation, and error messages all need to remain reliable for the people who already use and pay for it.

The product can also serve as a working example of how I design agent tools and content delivery. Someone can see the actual product, documentation, and workflow before deciding whether they need help with a larger knowledge base or operating system.

The second curve might become a content growth workshop. It might become consulting or a setup service. I still have unresolved delivery questions. How should a local system be upgraded after a customer changes its files? How much ongoing support turns a project into indefinite coaching? Which pieces can be standardized?

I do not want to launch all of those ideas at once.

The next useful step is to work with one or two real people, map their existing process, and deliver the smallest version that can run. Continued use and willingness to pay will tell me whether the work should become a workshop, a repeatable service, or another product.

§A Product Ceiling Is Not a Personal Ceiling

I used to tie the growth of a product too closely to my own progress.

When the product slowed down, I questioned whether the effort had been worth it. md2wechat has already left behind code, users, documentation, search entry points, and a practical body of agent-native engineering. It also moved me from building a formatting tool toward understanding content systems and industry delivery.

A narrower market does not erase those assets.

I still want to make md2wechat dependable. I no longer expect it to carry the entire growth story on its own. It can be a tested foundation for workshops, consulting, implementation work, or a future product.

I have not found the second curve yet. For now, I am keeping the main line alive, cutting low-value work, and using small deliveries to collect real evidence.

Some days that still feels difficult. I will keep moving, watch what happens, and try to stay level-headed.