Friction Was the Feature

The essay argues that generative AI has made writing and drafting so cheap that systems built on text as a proxy for effort or quality are losing their signal. Hiring, admissions, reviews, outreach, and claims all produce more polished output, but the extra volume makes judgment harder and often shifts institutions toward cruder filters.

The proposed response is to redesign processes around harder-to-fake proof, not more prose. The essay favors trials, structured exercises, timestamps, video evidence, and other forms of verification, while also suggesting small costs that discourage spam and using AI to compress work instead of inflating it.

Reading notes

#

• Friction used to make effort visible in jobs, emails, applications, and similar processes.
• AI lowers the cost of producing tailored text, so volume rises while judgment weakens.
• In the Freelancer.com example, stronger proposals became less informative after AI tools spread, and higher-ability workers were hired less often.
• The essay uses Goodhart’s law to describe how once-useful artifacts become targets and stop measuring what matters.
• The same pattern is described in product reviews, cold outreach, school admissions, and warranty or return claims.
• The loss of friction creates worse matching, more time spent on text, more complex systems, and less fairness.
• The proposed fix is to shift from prose to proof, add smart friction, and use AI to reduce text rather than multiply it.