The solution might be cancelling my AI subscription (Wilson)
David Wilson's confessional indictment of AI-assisted programming: a catalogue of ~70 projects built and abandoned, none useful, all enabled by the same tool that destroyed his focus.
Key Quotes
"almost none of this is useful and I don't want to maintain any of it"
The piece's governing admission. Wilson isn't an AI skeptic — he's a heavy user who looked at the output and found nothing worth keeping. The volume of creation is the evidence for the prosecution, not the defense.
"a thermonuclear ADHD amplifier"
The most quotable line, and also the most contested. Wilson sees AI as taking existing distractibility and weaponizing it: three screens, unrelated projects, zero commitment to outcome. The HN thread replies that for people with actual ADHD, AI can be the opposite — a tool that lets them finish before boredom kills the project. Both can be true; the difference is whether the user has a pre-existing relationship with focus they're trying to protect, or a deficit they're trying to fill.
"because the effort was removed, so was the commitment, and with the commitment the focus, and with the focus any meaningful product at all"
This is the structural claim, not just the anecdotal one. Wilson draws a causal chain: effort → commitment → focus → quality. Remove the first link and the whole thing collapses. His voice-to-blog experiment is the test case — removing the friction of writing produced output so bad he's embarrassed to have his name on it. This isn't an argument against keyboards; it's an argument about the relationship between cost and meaning.
"for as long as quality matters, handwriting can never be obsolete"
The most conservative claim in the piece, and the most interesting. Wilson isn't making a Luddite argument — he's making a definitional one. If quality is the goal, then the process that produces quality is the constraint. Remove the process and you remove the quality. AI tools are optimized for the wrong metric.
"a cheap reward with minimal input and no friction can only be a liability"
The thesis, distilled. Wilson is channeling the same mechanism Breaking the Spell of Vibe Coding identifies as "loss disguised as a win" — the slot machine that plays celebratory sounds while taking your money. The reward is real (dopamine, a working prototype) but the net outcome is negative (time burned, nothing maintained, skills atrophied).
"Almost every vendor and every tool intends to do exactly the opposite: more usage, more tokens, more output."
The incentives argument. ChatGPT's follow-up questions aren't a feature for you — they're a retention mechanism for OpenAI. Every interaction is designed to become the next interaction. Wilson is one of the few writers who names this honestly: the vendor and the user have opposed goals, and the vendor controls the interface.
Key Themes
- #concept Friction as Quality Control: The provocative idea that what we call "friction" in creative work is actually the mechanism that ensures commitment, focus, and quality. Removing it doesn't speed up quality work — it eliminates quality work's preconditions.
- #concept Thermometer vs. Amplifier: AI as an ADHD amplifier for people with baseline focus, but potentially a focus enabler for people who lack it. The same tool has opposite effects depending on the user's cognitive profile.
- #concept Vendor-User Misalignment: Every AI vendor's business model (more tokens, more sessions, more output) directly conflicts with the user's interest (solving problems and stopping). This is a structural conflict, not a UX problem.
- #person David Wilson: Engineer and blogger at thoughts.hmmz.org. Writes confessional tech critique from the position of a heavy user, not an outsider.
Critical Analysis
Wilson's piece is important precisely because he's not anti-AI — he's a power user who reached the end of the road. That's a more dangerous critique than any outsider's polemic. The vendor can't dismiss him as someone who doesn't understand the tools.
The strength: Wilson connects three observations that are usually kept separate — personal experience (I built 70 things and kept none), incentives analysis (vendors profit from my overuse), and a theory of quality (friction isn't a bug). The essay earns its pessimism; it's not a hot take, it's a post-mortem.
The weakness: The prescription is non-existent. "Curtail use" is a personal resolution, not a strategy. The piece has no theory of good AI use — it's entirely about the failure modes. This is honest (Wilson admits he has no solution) but limits the essay's utility. Compare to Agent Coding Workflow, which catalogues precisely the set of practices that separate productive AI use from time-wasting. Wilson describes the disease but not the cure.
What's missing: The distinction between building and engineering. Wilson's 70 abandoned projects are all greenfield builds — quick scripts that became unmaintainable apps. He never describes using AI to work on an existing codebase, to understand a complex system, to debug a production issue, or to learn a new domain. These are the use cases where AI's value is highest and its addictiveness is lowest. The essay conflates "AI-assisted coding" with "AI-assisted greenfield projects," and they're not the same thing.
The neurodiversity dimension: The HN thread complicates Wilson's thesis in ways the essay doesn't anticipate. If AI helps people with ADHD finish things for the first time, the "thermonuclear ADHD amplifier" framing is too narrow. The variable isn't the tool — it's the user's pre-existing relationship with sustained attention. This is the most interesting open question the piece raises: can the same tool be simultaneously a focus-destroyer for some and a focus-enabler for others, and if so, what determines which camp you land in?
See also: The solution might be cancelling my AI subscription (Willison) for the companion response.
Sources: The solution might be cancelling my AI subscription Last updated: 2026-06-15