Confessions of a Chronic Blog Procrastinator

I've been maintaining a tech blog for over two years, and the backend data is brutally honest: an average of 1.3 posts per month, while my plan was one post per week. Where's the gap? Not in writing speed, but in everything that happens before writing.

Let me break down my old process for writing a single blog post:

• Topic selection: 45–90 minutes. Digging through bookmarks, notes, and topics discussed in group chats, repeatedly second-guessing myself with "would anyone actually read this?";
• Research to validate the topic: 30 minutes. Searching whether others have written about it and how well;
• Outlining: 60–90 minutes. Getting stuck at second-level headings, often tearing up half-finished outlines and starting over;
• Actual writing: 90 minutes.

The first three steps often added up to more than 3 hours, and it was all low-certainty labor—after investing three hours, the conclusion might well be "forget it, I'm not writing this." Worse, the frustration of getting stuck during topic selection and outlining kills the motivation to write outright. The root cause of procrastination isn't laziness—it's the sky-high startup cost.

After AI Took Over the First Three Steps

My workflow now looks like this:

Step 1: Feed it raw material. Take the past two weeks of tech notes, bug-ridden war stories, and questions I've answered in group chats—no matter how fragmented—and throw them straight at the AI. No organizing, no polishing, just raw material dumped in.

Step 2: AI does the initial topic screening. Have it output 5–8 candidate topics, each with: target readers, pain points addressed, expected structure, and most importantly—why this topic suits me and why it doesn't. It's more honest than I am; it'll flat-out say "this topic is heavily covered already, unless you have exclusive data."

Step 3: I pick one, AI produces three outline versions. Have it generate an outline in three different structures (tutorial-style, postmortem-style, argument-style). I pick one or stitch them together, then have it expand to third-level headings, marking the factual material needed for each section.

Step 4: I write following the outline. AI can't help much here—first-person technical details and real experiences are things only I have. But with an outline in hand, writing time dropped from 90 minutes to about 60, because I'm no longer figuring out structure while writing.

Before and After, in Numbers

2–3 hours saved per post. At a target of four posts per month, that's at least one full workday saved monthly. Publishing frequency went from 1.3 posts per month to over 3—when the startup cost drops, writing just happens naturally.

My Prompt Template

I've iterated on this template a dozen times. Copy and adapt it directly:

你是一位技术博客选题顾问。我会提供我的技术笔记和零散想法。

请完成以下任务：

1. 从素材中提炼 5 个博客选题，每个选题包含：
   - 标题（不用夸张修辞，描述准确即可）
   - 目标读者（具体到角色，如"刚接手运维工作的后端开发者"）
   - 核心痛点（读者为什么需要这篇文章）
   - 一句话价值主张（读完能获得什么）

2. 对每个选题做诚实评估：
   - 该话题在网上的覆盖程度（高/中/低）
   - 我的素材是否足以支撑（指出缺什么事实）
   - 适合我的写作风格的理由 / 不适合的原因

3. 我选定一个选题后（我会回复编号），请输出三版大纲：
   - A 版：教程式（step by step）
   - B 版：踩坑复盘式（问题→尝试→结论）
   - C 版：观点论述式（现象→分析→建议）
   每版到二级标题即可。

4. 我选定大纲版本后，把该版补全到三级标题，
   并在每一节下标注「需要我提供的事实素材清单」。

规则：
- 不要编造我没有提供的数据、案例或经历
- 评估选题时宁可泼冷水，不要鼓励我写没把握的内容

我的素材如下：
【在此粘贴你的技术笔记 / 聊天记录 / 踩坑记录】

Rule 4 is the most important: AI handles structure and judgment, but the facts must come from me. Hold this boundary, and AI-generated outlines won't carry that "manufactured confidence" that leaves you discovering pitfalls everywhere when you actually write.

A Few Practical Tips

• The process matters more than model choice. I've switched between three models running the same workflow, and the difference in results was far smaller than the difference between good and bad prompts. Spend your energy polishing the template. For model options and API costs, refer to the official pricing page.
• Don't skip the "cold water" step. AI agreeing with "this topic is great" is worthless. Explicitly require it to give counterarguments, and topic quality will improve noticeably.
• Third-level outline headings are the sweet spot. Detailed enough to start writing immediately, but not so detailed that AI writes the body for you and you lose your personal voice.
• Batch your topic selection. I now do a "feed + select" session every two weeks, building a topic pool. When it's time to write, I just pull from the pool—saving even those 15 minutes of topic selection.

Final Thoughts

What's truly valuable in the writing process is your technical judgment and real experiences, while the grunt work of topic screening and structure building can be entirely delegated to AI. Trading three-plus hours for one and a half means, for an indie developer, one more piece of search-traffic-generating technical content per month—with considerable long-term compounding.

If you want to run this kind of workflow at scale, turning topic selection, outlining, and draft polishing into a stable API pipeline, check out Thistoken's API service—unified access to multiple models, with registration here: https://api.thistoken.ai/register

---

Ready to try it yourself? Sign up at https://api.thistoken.ai/register to get your API key and start building.