I don't want to read what you didn't write

Colin Breck criticizes the growing habit of using AI to generate design proposals, business plans, tickets, pull request summaries, meeting notes, and even personal messages, arguing these outputs are often unreadable because they lack authorial context, voice, and vulnerability. Machine-generated summaries frequently convert completed work into exhaustive, decontextualized descriptions that force readers to parse details without guidance on value, urgency, or where input is needed. Readers respond poorly - survey data show most stop reading AI-scented pieces and overwhelmingly prefer an author’s own imperfect voice - and AI-mediated messages can feel impersonal, dispassionate, and disjointed, eroding the social contract between writer and reader.

Breck distinguishes harmful uses from productive ones: AI excels as an assistant that verifies facts against source code, configurations, logs and metrics; automates citations; finds grammar and notation errors; and draws technical diagrams, all of which sped his LaTeX paper writing and even produced a useful abstract. He suggests targeted interventions - applying controlled English standards like ASD-STE100 for technical documentation and using detectors such as Pangram to flag AI-authored text - to curb lazy, contextless writing. The central claim is firm: AI should augment human judgment and preserve narrative, nuance, and uncertainty, not substitute for authentic human-crafted prose.