If You Want Human Attention, Show Human Effort: The #1 HN Rule and Where It Breaks

Summary

A two-minute post, published June 11, reached the top of Hacker News within two days: 1,501 points, 460 comments. Tom Bedor’s rule is a single line. If you are asking for human attention, demonstrate human effort.

His argument is plain. A growing share of debug investigations, documents, and code is written by robots, and as engineers spend more of their day reading AI text, fatigue sets in. He names the source of that fatigue precisely: if I can have a robot say something, so can you. Posting undigested AI output as though it were your own writing reads as inconsiderate. The moment that lit the idea was concrete. He proposed a design, a teammate had an AI critique it, and forwarded the document with a disclaimer: I did not read this, so it might not be entirely accurate. His reaction is the real engine of the piece. If reading this was not worth your time, why is it worth mine.

So he adopted a discipline. He still sends useful AI content to teammates, but he labels what is AI-generated and adds his own commentary, and for code review he always reviews his AI-generated code first. Attention was scarce before AI and is scarcer now. Labeling and showing effort is consideration for teammates, and keeps a touch of humanity in the work.

The post landed less for novelty than for voicing an unspoken irritation shared by a whole cohort of engineers. Topping HN and holding up under scrutiny, though, are two different things. Across the 460 comments, the real disagreement is sharper than the headline.

The debate

On the surface everyone nods. Read further and the comments split into two camps that cannot both be right, fighting over one question: when you decide whether something is worth reading, is the test the effort that went in, or the quality that comes out?

The quality camp brings the heaviest fire, led by a highly upvoted rebuttal from commenter thaumasiotes. The rule, he argues, is just the labor theory of value applied to documents, and value does not track how hard something was to make. If I do something easy for me that is useful to you, you still want it. If I do something hard, the difficulty implies lower quality, so it may be worth less. The same holds for machine-generated text: if it is valuable, read it; if not, do not; the tool is irrelevant. Another commenter, mapontosevenths, is blunter: what I want is quality, the sweat of your brow is just salt water, so use whatever tool does the job and own it. A third, pevansgreenwood, invokes Pascal (I made this letter longer because I had not the time to make it shorter) and adds that respect is measured by whether the message serves the recipient’s needs, not by how it was produced.

The effort camp does not meet that head-on. It slips in through a crack. The reply from rodonn is the hinge of the whole debate: the problem is that I do not know whether a document is useful until after I read it. Because value cannot be judged in advance, I want the sender to first show they invested effort proportionate to the time they are asking of me. Effort is not the value; it is a deposit against it. Aldipower makes the abstraction concrete with a real case: he runs a product that talks to LLMs, and every day he gets support emails written by users’ LLMs, full of hallucinations, that a single human read-through would have caught.

A third group moves the battlefield off quality entirely. nlawalker argues the rule misses the point and should read: if you are asking someone to assume accountability, demonstrate human effort. In his experience, people making these requests do not want your attention at all; they want you on the hook for something. Jimmc414 echoes the legal analogy: the human submitting the work, not the AI, is responsible for its quality. That thread matters, because it quietly drags the debate from etiquette toward accountability.

Who’s right

Both camps catch something real, and both miss a piece. Fit the pieces together and the picture closes.

The quality camp is right about the endpoint. What you ultimately care about is usefulness and accuracy, and nobody should applaud effort for its own sake. A useless report that took ten hours does not become worth reading because it was slow. Treating effort as the goal does slide into the labor-theory trap.

But the quality camp misses the timing problem rodonn names: quality is verified after the fact, while attention is paid before it. You have to read the thing to learn whether it was worth reading. That is a bet under information asymmetry. Before AI, the odds on that bet were tolerable, because writing a long document was itself a filter, and people willing to spend that effort usually had something to say. AI removed the filter. With production cost near zero, length, structure, and polish no longer prove that thought went in. So effort degrades from a goal into a signal: when you cannot verify quality in advance, whether someone invested proportionate effort is the only forward indication of quality you get.

The cleanest version of Bedor’s rule, then, is not effort has value. It is the sharper line from commenter glennericksen: spend more effort producing something than it takes someone else to consume it. That appeals not to labor value but to reciprocity. It explains why two AI-written documents land so differently, one appreciated, one resented: the difference comes down to who pays the bill, not whether AI was used. Shifting the full cost of verifying, correcting, and understanding onto the recipient is the line that gets crossed.

The effort camp needs tightening too. Setting the test to effort invites gaming through I spent a lot of time, and nlawalker’s reminder cuts deeper: often the other party does not want your attention at all, only your signature. There, showing effort is not enough; the real red line is accountability. Whoever submits it owns it, and AI is not a disclaimer.

My read: Bedor points the right direction with soft wording. What should survive is not the phrase show effort but the reciprocity law and the accountability law behind it. Effort is the means; reciprocity and ownership are the point.

Why it matters

This deserves a builder’s attention less as workplace etiquette than as a sign of something being repriced: when the cost on the production side collapses toward zero, attention on the consumption side becomes the only resource still scarce in the system.

Many collaboration workflows assumed a premise that has now failed: production is costly, so people do not produce frivolously. Code review worked because writing a PR limited the number and size of PRs. swiftcoder on HN puts it cleanly: review never scaled to prolific humans and definitely cannot scale to agents, and the things you would need to safely drop PRs (auto-formatters, linters, end-to-end tests, continuous deployment) are exactly the guardrails that keep an LLM in line. niuzeta gives a live example: a prolific coworker who fully embraced Claude flooded the team with AI-generated PRs, and six months later complained at standup that his PRs went unreviewed. No one avoids them on purpose; they are just hard to review. When production cost goes to zero and verification cost stays, the bill eventually settles as silence.

For teams that write email, PRs, and design docs with AI, the usable boundary is clear. One, label the AI-generated parts, not as a moral gesture but so the reader knows where verification responsibility sits. Two, pay the understanding and correction cost yourself before you ask for someone’s time, at minimum enough to stand behind the content. Three, write the reciprocity law into team norms: producing should cost more effort than consuming, or you are overdrawing the team’s attention budget. None of this is anti-AI. It simply reprices attention after AI erased the cost of production.

What to ignore

Half the noise in the thread can be skipped. It scratches the emotion, not the problem.

Skip the revenge genre above all. Some suggest the most pedantic review bot to grind the offender down, some want Claude to adversarially review the offender’s PR into oblivion, some just say fire him. Satisfying, but it escalates the arms race, burning more compute on both sides to cancel each other out, which is the opposite of reducing wasted attention. The real fix is the plain one from AussieWog93: talk to the person directly. People who make this mistake are usually unaware of why it is wrong, not malicious.

Watch the other extreme too, the allergy TFNA worries about: because long text reads as AI-generated, people start avoiding anything with depth or length. Equating short with sincere is its own laziness, and some things need a few solid paragraphs to say at all. What deserves penalty is the undigested forward, not the word count.

Finally, do not get pulled into the meta-debate over whether AI is even good enough to read. Some say AI quality is long past the complaint stage, some say it never deserved it. That argument has no endpoint and does not help, because Bedor’s rule does not depend on whether AI is good. Even if the AI wrote it perfectly, dumping output you have not vetted onto someone else to verify still shifts the cost. The thing to hold onto is the reciprocity law.

Sources

1. If you are asking for human attention, demonstrate human effort (Tom Bedor) / blog
2. If you are asking for human attention, demonstrate human effort (Hacker News discussion) / hn

No official primary source available; this analysis is based on reliable secondary reporting (named outlets, cross-confirmed).