A new piece making the rounds on Hacker News attempts to sort the dumpster fire of AI discourse into something resembling order, and it's equal parts brutal and cathartic for anyone who's been drowning in this stuff for the past two years. The taxonomy, published at iceworks.cc, identifies five distinct archetypes currently wrecking havoc across repositories, comment sections, and corporate strategy meetings.

The Never-Clankers

First up: the 'Never-Clankers,' who apparently think the solution to AI-generated code is to lynch any contributor who's ever touched an LLM. The piece rightly calls out how absurd this position is, pointing out that even Linus Torvalds and Fabrice Bellard have acknowledged large models can be useful for low-level work. 'That position is a house on a broken foundation that is sinking in to the swamp,' the author writes. These folks get labeled Luddites, though the piece argues that's historically inaccurate—real Luddites were concerned about labor rights, not just ejecting anyone who thought a sewing machine was nifty.

Vibe Coders: The Real Problem

Then there's the 'Vibe Coders,' and this is where things get spicy. These are the developers so brain-rotted from LLM overuse they don't even read code anymore—just copy-paste errors back into prompts until something works. The author doesn't sugarcoat it: they're either terrible coders, completely fried, or playing corporate games by letting AI write garbage while collecting paychecks. Projects like MemVID get called out specifically for having documentation so wildly out of sync with actual software behavior that nobody knows what's true anymore. 'Bad devs cope about moving too fast to document,' the piece notes. 'Worse devs document everything wrong.'

The Reasonable Takes (Finally)

Not everyone gets dragged, though. The 'Utility Minions' category covers unskilled users leveraging LLMs for one-off scripts—like using Gemini at work to reboot routers—without pretending they're software engineers. The author thinks this is actually fine. These people are honest about what they're doing and the $10 in token fees gave them something they couldn't otherwise obtain. There's a genuine argument here that power-user tools from the 1990s era—the Warcraft 3 trigger editors, Macromedia Fusion event sheets—did more for 'on-ramping' non-programmers than current AI hype.

The Demings: Doing It Right

The most sympathetic category goes to 'The Demings,' named after W. Edwards Deming and his '100% inspection' philosophy. These are the utilitarians who say sure, use whatever bots you want—but you're personally responsible for explaining every line when it hits production. Your name's stapled to that release, so you'd better know what you're signing off on. The piece argues this is the sustainable approach because ultimately, 'looking at the quality of code is the right metric regardless of how it got made.'

Artificial Idiot Managers: The Real Villains

And then there's the category nobody should want to be in: 'Artificial Idiot Managers,' who quote Marc Benioff about being 'the last generation to manage only humans' without a hint of irony. These are the execs with an actual fetish for replacing employees—not using AI as a tool to reduce tedium, but dead set on firing their entire team and collecting industrial revolution rewards before actually shipping anything superior. The author points out historical context: sewing machines won because they did consistent work faster and workshops could clear backlogs. AI isn't yet delivering the product—executives just want to rush to the rewards.

Key Takeaways

  • Vibe Coders are arguably more dangerous than Never-Clankers—they 'write bad code faster' at scale
  • Utility Minions represent acceptable, honest LLM use that actually solves real problems
  • The Demings approach (full ownership regardless of tool) is the only sustainable path forward
  • AI hype has attracted grifters on both sides: execs chasing replacement fantasies and anti-AI cultists

The Bottom Line

This taxonomy cuts through the noise because it's written by someone who's clearly been neck-deep in actual projects dealing with this mess. The categories aren't flattering, but they're accurate—and if you're not seeing at least one of these archetypes in your daily work, you probably work alone.