I paid people to try and follow my README

TEXT START: Developers suck at writing documentation.

The Dissection

This is a field report on usability testing disguised as a defense of human feedback. The author turns a README from a static artifact into an executable interface: recruit unfamiliar users, watch them fail, capture their confusion, revise, and repeat.

The €25 is not charity. It purchases unfiltered error telemetry that the author’s own expertise has erased. The article correctly identifies tacit knowledge, bad sequencing, unexplained terminology, and authorial assumptions as the real failure points. Its anti-LLM argument is narrower: synthetic user simulation cannot yet reproduce the full signal of live confusion, conversation, frustration, and improvisation.

The Core Fallacy

The article confuses human usefulness with human indispensability.

Real people may be superior test subjects for this task. That does not make human labor structurally necessary across the economy. The workflow is a temporary verification niche: humans expose failure modes while the underlying software, documentation, testing, and feedback analysis continue moving toward automation.

“People are brilliant” is an epistemic observation, not an economic defense. The fact that human confusion improves a product does not preserve the mass employment-to-wage-to-consumption circuit. It merely proves that some human judgment remains valuable during the transition.

Hidden Assumptions

• The recruited testers are representative of actual users.
• Small-scale interviews reveal the most important failure modes.
• The author can afford repeated manual testing and revision.
• Users can articulate confusion accurately while screen-sharing.
• Live emotional feedback contains signal that automated telemetry cannot extract.
• The product’s main adoption barrier is documentation rather than software quality, trust, or distribution.
• Volunteer or low-paid testers remain available as a scalable resource.
• Human-centered usability work will remain cheaper or better than increasingly capable automated alternatives.
• Users retain enough technical autonomy to install and operate the software themselves.
• Better documentation expands a viable user base rather than merely polishing a shrinking human-facing niche.

Social Function

Primary classification: partial truth and transition management.

The practical advice is sound. It reduces friction during the lag period when software still depends on heterogeneous human users, imperfect interfaces, and trust. It is also a form of verification arbitrage: buying cheap human confusion and converting it into a more valuable automated product experience.

The sentimental defense of real people adds a small layer of ideological anesthetic. It treats the continued usefulness of human perception as evidence of continued human centrality. That inference does not follow. A system can pay a few people to identify defects while eliminating the broader population’s economic necessity.

The Verdict

The article is correct at the interface and blind at the system level. Human testing can make a README materially better, but it is a lag defense, not a counterforce to obsolescence. Under the Discontinuity Thesis, this is a narrow servitor or transition-intermediation niche: valuable because automation has not fully closed the loop, not because humans have regained structural power.

The README may become easier to follow. The mass labor circuit remains headed for the morgue.