#llm-skepticism

38 pages search “llm-skepticism” ↳ all tags

Wiki 38

  • "No AI" Statements Are Necessary and Useful James Zhan's rebuttal to the "just write well and readers will know" defense — handmade labels predate AI, detection is broken in both directions, and the marker is a stance
  • A Voice From Nowhere zanlib on Leo XIV's encyclical: generated work copies the form of thought, not the path under it
  • AI detector arms race Surface-based AI detection fails in both directions, the tells change every model release, and the stable equilibrium is writers degrading their work to pass
  • Amy J. Ko blog UW Information School professor (HCI, CS education) who writes Bits and Behavior on Medium, mostly LLM-skeptical takes on programming and learning
  • Anthony Moser (anthonymoser.github.io) Anthony Moser's GitHub Pages blog; "haterdom" essay series critiquing AI, with community-translated editions
  • Anti-LLM Discourse The 2025-2026 cluster of public anti-AI / anti-LLM writing — its sub-genres, who occupies which slot, and what each is actually arguing
  • Antman Does Software (Anthony Manning-Franklin) Anthony Manning-Franklin's blog at antman-does-software.com on software craft, learning, and AI skepticism
  • clankerfree (Software for Humans) clankerfree.org lists open-source projects and hosts that self-certify using no generative AI, under a three-rule inclusion policy
  • Code review throughput limits The empirical caps on effective code review (~400 LOC/h, ~1h sessions, no parallelism) and why they put a hard ceiling on "just review the AI's output"
  • Credibility as the slop test Whether AI-assisted text is slop can't be judged from its surface — the only working test is whether the author would read it aloud, unashamed, staking their credibility
  • Dan Luu on AI Coding, Testing, and Variance Dan Luu applies a CPU-verification background to agentic coding — testing-heavy no-review workflows, the meaninglessness of single-number model benchmarks, and working around agent failure modes
  • Defining AI Psychosis, Part 2: Prolific AI Psychosis Psychiatrist Jeff Clark defines prolific AI psychosis as heavy AI output with little real value and a lost ability to judge one's own work
  • Don't Call Yourself an Artisanal Programmer purplesyringa argues the "artisanal" label hands the word engineer to AI users, and proposes "AI-free software engineering" instead
  • How I Feel About AI (Zwinkau) Andreas Zwinkau names six feelings about AI and concludes it looks positive at the technological level but bleak for society
  • Human-made disclosure The declare-human marker — a footer line, signature, or store label asserting no AI was used; unverifiable by design, valuable as a stance rather than as evidence
  • I Am An AI Hater (Anthony Moser) Moser's August 2025 essay rejecting polite AI discourse, cataloging documented harms, and framing AI as fundamentally fascist
  • I Think You Might Be Fooling Yourself With AI louwrentius: you can't measure your own AI productivity gain, and the price you pay is subsidized
  • I Will Never Use AI to Code (Manning-Franklin) Anthony Manning-Franklin's nine-reason refusal, stitching personal-craft, skill-decay, economic, and ethical arguments into one declaration
  • I'm Going Back to Coding by Hand (Ask HN) An HN user abandons six months of Claude-built features on their own app after realizing they no longer knew how the new code worked
  • It Matters Who Teaches You (Gruber) Bastian Gruber on why an LLM can explain but not teach, borrowing Susan Piver's point that a teacher brings a lineage and knows the student
  • James Zhan (journal.james-zhan.com) James Zhan's personal journal — writing, typography, Mac software, and the case for declaring human authorship; footer reads "100% written by James, not AI"
  • Know Thine Enemy (Ko) Amy Ko's 3-month experiment using Claude Code as an LLM skeptic — it worked, and cost her focus, joy, money, and values
  • LLM output variance Run-to-run and task-to-task variance in LLM/agent output is high enough that small-sample comparisons and single-number benchmarks support almost any conclusion
  • Measuring AI Coding Productivity The recurring methods failures that make most claims about whether AI coding tools help unreliable — proxy metrics, missing controls, novelty, selection, systems confusion
  • Memorizing Session Transcripts Isn't Useful theahura found zero SWE gain from agents searching their own past transcripts, and explains why
  • No, AI Is Not Conscious (Ted Chiang) Chiang argues LLMs are sentence-continuation machines with no subjective experience, so they can be neither conscious nor moral agents
  • Programming still sucks (stvn) Stvn's drunk-at-a-birthday-party essay reframing the AI-took-our-jobs question — greed did, AI is the mask, and the company is held up by a 55-year-old named Sara who lives below decks
  • READMENOT — a marker for code not meant for humans William Woodruff's proposal for a READMENOT file that honestly signals a project is slop — not meant for unwitting human reading
  • Reviewing AI Code Thomas Depierre's empirical case that "just review the AI's code like an intern's" collapses against what research says about code-review throughput and reviewer overconfidence
  • Slop-marker convention The idea of an explicit, machine-and-human-readable marker declaring code (or text) as slop — defined by its production process, not its surface
  • stvn.sh stvn's writing site at stvn.sh — first-person essays on tech work, AI corporate culture, and institutional knowledge
  • Ted Chiang Science-fiction writer and essayist; recurring critic of LLM hype and anthropomorphism for The New Yorker and The Atlantic
  • The LLM Critics Are Right. I Use LLMs Anyway Jeremy Theocharis on the dissonance of agreeing with nearly every anti-LLM argument while spending $10k/month on tokens — and the workflow patterns that make it defensible
  • theocharis.dev Jeremy Theocharis's blog on LLM-assisted work, local-first software, and engineering practice; writes with LLMs about using LLMs
  • Thomas Depierre (Musings about software) Elixir and DevOps consultant blogging at softwaremaxims.com; evidence-first takes on software practice and LLM skepticism
  • Twelve Ways to Be Wrong About AI-Assisted Coding Greg Wilson's catalog of 12 measurement errors in studies of whether AI coding tools work, each mapped to a known research-methods failure
  • Working around agent failure modes is the skill Most of the skill in using coding agents is knowing their specific failure modes and building around them — a skill that decays as labs fix those failures
  • Zig's Anti-LLM Policy and the Bun Fork (Simon Willison) Simon Willison on Zig's strict no-LLM contributor policy, Loris Cro's "contributor poker" framing, and Bun's parallel codegen fork that won't be upstreamed because it was AI-assisted