362 points HN explosive article uses personal experience to warn: AI subscription is destroying your concentration

Core conclusion

A 362-point HN personal essay "The solution might be cancelling my AI subscription" triggered a heated discussion with 229 comments. The author David is not anti-AI, but has personally experienced how AI fragments the attention of knowledge workers - low-friction AI tools create a lot of "useful garbage" rather than real output.

This is an important warning sign for content operators who rely on AI agents and automation tools: a tool is only as good as not what it can do, but how you use it. **

Key Points

• Source of event: Hacker News personal blog "the solution might be cancelling my AI subscription" -Affected objects: All developers, independent makers, and content operators who rely heavily on AI coding/content tools
• Core point of view: AI tools eliminate all friction, but friction is precisely the prerequisite for concentration and product quality

Background: What does this explosive article say?

On May 31, a developer named David published a short personal reflection on his blog thoughts.hmmz.org. He lists the 50+ projects he’s built with AI—a speech recognition system, an email archiving tool, a Jellyfin desktop client, a news site—and then confesses: **Except for being a SaaS, almost all of them are useless, and he doesn’t want to maintain any of them. **

"Except for the SaaS, almost none of this is useful and I don't want to maintain any of it."

David said that every Claude session started with a simple request to "write a script to handle AI removes the friction of building, but also removes the incentive for commitment, focus, and ongoing maintenance.

He described AI as a "thermonuclear ADHD amplifier" and observed the same phenomenon happening to all of his adult friends - three screens open at once, working on completely unrelated "projects" with little hope of maintenance.

"Slopping out a 10,000 LOC untested Python/JS mess in 5 minutes helps nobody."

HN community’s polarizing reactions

The post prompted 229 comments on HN, with the community sharply polarized.

The support camp believes the problem is real. Many developers shared their own similar experiences:

"AI reduces the time cost of making an initial product, bypassing the need for real commitment, investment, intense interest, and dedication—the very things that keep a project alive."

One older developer put it in terms of time scarcity: "I'm in my 50s. I was happy to chase down these rabbit holes in my 20s, but now time is so scarce that anything that makes me spend time on things that aren't really important is a strong negative."

The opponent camp believes that the problem is not AI, but the way it is used and personal discipline:

"AI was the opposite for me - I finished my side project for the first time because I was able to actually get it running before I got bored."

"Your car can go 100+ mph, but you shouldn't be driving at top speed on street corners or around your neighborhood. The same goes for LLM. Practice restraint and focus, and LLM will make you more efficient."

There is also a middle path emerging in the community that is worthy of attention - using AI in specific areas can bring real efficiency gains, and the key is to distinguish between "exploratory use" and "productive use."

Core Conflict: Is eliminating friction a good thing or a bad thing?

This article gets to the core paradox of AI tools.

The selling point of AI has always been "elimination of friction" - less learning costs, faster prototyping, and lower development thresholds. This all sounds like a good thing. But David points out: **Friction is not just a barrier, it’s also a filter. **

The author cites Cal Newport's concept of "pseudo-productivity" to explain this phenomenon: digital productivity tools (including AI) make individual tasks faster, but make knowledge workers busier, more distracted, and less productive overall.

Inspiration for heavy users of AI tools

As a team running an AI content site, the warning from this article is very direct. The following three items can be implemented immediately:

1. Setting friction points for AI usage: Not all tasks are worth accelerating with AI. Reserve "AI-free time" for tasks that require deep thinking, and don't open the chat box when writing blog posts/making decisions.

2. Distinguish between building projects and throwing prototypes: AI is good at throwing prototypes, but not good at maintenance. Before starting a new project, ask yourself "Do I still want to maintain this in three months?" If the answer is no, keep only minimal experimental code and don't invest 10 hours in making it perfect.

3. Assign AI permissions by role: Content generation uses AI to accelerate drafts, but editing and strategy decisions must be done by humans. This is exactly the content production strategy of the WayToClawEarn team - AI produces the first draft and humans complete quality control.

Related further reading

• AI 编码贬值，领域知识升值
• AI 代码质量 vs 速度：当"快"变成陷阱
• AI 行业冰火两重天

Tool entry

Tool names that appear naturally in the text: Claude, ChatGPT, Claude Code, Codex, Cursor, OpenAI