What Happened
Rob Mealey published an essay on Substack titled "AI is a Mood, Not a Method," arguing that artificial intelligence is better understood as an ideological aspiration than a technical discipline. The piece traces a repeating pattern from the 1956 Dartmouth Conference through expert systems to today's large language models: each cycle begins with predictions of machine consciousness, then quietly ends with the useful tools getting rebranded as something else (search algorithms, enterprise software, compilers) while shedding the "AI" label.
The essay leans on Marx's concept of alienation to frame three waves of technological abstraction: factories alienated labor, social media alienated attention, and AI now threatens to alienate cognition itself. Mealey calls LLM training data "the greatest enclosure of the commons in human history," pointing to every Reddit thread, coding debate, and advice post scraped without compensation, plus the underpaid workers in developing countries who label training data for pennies.
He names the beneficiaries directly: venture capitalists who need inflated valuations, executives who need to justify infrastructure spending, cloud providers who need compute demand, and politicians who find "arms race" narratives convenient.
The essay highlights two counterexamples of AI done right: Anat Caspi's OpenSidewalks project and Ted Underwood's digital humanities work, both of which extend human capabilities rather than promising to replace human thinking.
Why It Matters
If you use AI tools daily, this essay is worth sitting with, even if you disagree with half of it. The cyclical pattern Mealey identifies is real and observable. Capabilities that were "AI" five years ago are now just "features." Autocomplete, spam filtering, recommendation engines - all shed the AI label once they became mundane. Understanding this pattern helps you evaluate new releases with clearer eyes.
The hidden labor point hits harder the more you think about it. The models we use every day were trained on freely contributed human knowledge. That is not a moral judgment on using them. It is a fact worth holding in mind when companies claim these systems are purely technological achievements.
For practitioners, the most useful takeaway is the distinction between tools that augment your thinking and tools that promise to replace it. The former tend to deliver. The latter tend to disappoint and get rebranded.
Our Take
Mealey's framing is sharper than most AI criticism because he is not arguing that the tools are useless. He is arguing that the mythology around them is deliberately misleading. That tracks with what we see daily reviewing AI products. The tools that work best are the ones with the most modest claims - transcription, search, summarization. The ones that promise "autonomous agents" and "artificial general intelligence" are usually the ones that need the most hand-holding.
The essay overstates the alienation argument in places. Using Claude to draft an outline is not the same as a factory extracting surplus labor. But the core observation - that "AI" functions more as a marketing category than a scientific one - is hard to argue with. Every vendor we review calls themselves AI-powered now, whether they are running a frontier model or a regex with a chatbot skin.
Read the essay. Disagree with the parts that deserve disagreement. But take the pattern recognition seriously. It will make you a better evaluator of every tool launch that crosses your feed.