Bayesian Persuasion

Sources: How to Write with AI Without Producing Slop

4 min read

This document details the sources for claims made in the article. Claims are organized by the section in which they appear.

Garbage In, Garbage Out

Claim: The phrase “garbage in, garbage out” dates back to the late 1950s, and George Fuechsel, an IBM instructor, popularized it.

Source: Wikipedia, “Garbage in, garbage out” — The earliest recorded use appears in a 1957 newspaper article about US Army mathematicians. Most sources credit George Fuechsel, an IBM 305 RAMAC instructor, with popularizing the phrase between 1958-1959.

Claim: Google Flu Trends overestimated flu prevalence by more than 50% and was quietly discontinued.

Source: TIME, “Google Flu Trends Failure Shows Drawbacks of Big Data” — The system overestimated flu prevalence by more than 50% in 2012-2013 and was discontinued in 2015. Harvard’s Gary King called this “big data hubris.”

Additional context: Data Collaboratives case study on Google Flu Trends

Claim: Gartner predicted 60% of big data projects would fail; the actual rate approached 85%.

Source: INFORMS, “Why most big data analytics projects fail” — Gartner predicted in 2015 that 60% of big data projects would fail; analyst Nick Heudecker later admitted they were “too conservative” and the actual rate approached 85%.

What AI Does Well (That I Don’t)

Claim: A 2023 study of 758 BCG consultants found speed increased by 25%+ and performance improved by 40%+ for tasks within AI’s capability boundary, but consultants using AI were 19 percentage points less likely to produce correct solutions for tasks outside that boundary.

Source: Stanford Digital Economy Lab event summary, “Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality” — This was a BCG/Harvard study of 758 consultants. The “jagged technological frontier” concept comes from this research.

My Process (Section 8: Review It Thoroughly)

Claim: CNET published 77 AI-generated articles and had to correct 53% of them.

Source: Engadget, “CNET had to correct most of its AI-written articles” — CNET corrected 41 of its 77 AI-written articles (53%), including basic compound interest calculation errors.

Additional source: CNN, “Plagued with errors: A news outlet’s decision to write stories with AI backfires”

Claim: The fluency of AI writing reduced editorial scrutiny (Hany Farid quote context).

Source: Neuron Expert analysis of CNET incident — UC Berkeley’s Hany Farid identified that the compelling, fluent nature of AI writing may have caused editors to “exercise less scrutiny” compared to human-generated content.

What I’m Trading Away

Claim: Leslie Lamport’s quote about thinking and writing: “If you’re thinking without writing, you only think you’re thinking.”

Source: Goodreads attribution to Leslie Lamport (Turing Award winner)

Claim: Paul Graham warned that a world divided into those who write and those who don’t is really a world of “thinks and think-nots.”

Source: Paul Graham, “Writes and Write-Nots” (October 2024)

Additional Paul Graham essay on writing and thinking:

Additional Research Sources (Background Context)

These sources informed the research synthesis but are not directly quoted in the final article:

On Context Engineering and Prompting:

Andrej Karpathy on context engineering (June 2025):

Anthropic’s guidance on context engineering:

Anthropic case study showing 20% accuracy improvement with prompt engineering:

Chain-of-thought prompting research (Wei et al., 2022):

Survey finding 83.7% of users agree clearer prompts lead to better results:

Prompt Engineering Guide:

On AI Content Failures:

Sports Illustrated fake author scandal — fabricated identities with AI-generated headshots, CEO fired, stock dropped 22%:

  • Multiple news sources covered this in November 2023

Gannett LedeAI experiment — broken templates like “WINNING_TEAM_MASCOT”:

On Human-AI Collaboration:

MIT Sloan on human-AI collaboration patterns:

Ethan Mollick’s “Co-Intelligence” book coverage:

Steve Jobs “bicycle for the mind” concept (1981):

On Writing and Thinking:

Joan Didion’s “Why I Write” (1976):

WordRake on writing as thinking for knowledge workers:

Steven Sinofsky on “If Writing is Thinking…”:

Academic research on writing and critical thinking (Quitadamo & Kurtz, 2007):

Generated as part of the AI-assisted writing process. All major claims should be independently verified before publication.

Graph View

Backlinks