The Rise of the Bullshittery

Marius's long, exasperated take on why the modern professional economy has stopped rewarding people who do the work and started rewarding people who look like they do it. The frame is Harry Frankfurt's 1986 On Bullshit — the bullshitter is dangerous not because they lie but because they're indifferent to whether what they're saying is true. The argument is that platforms, incentives, and now LLMs have made indifference cheap, scalable, and on average more profitable than care.

The mechanism

The unspoken career contract — learn a thing, do it well, build a slow reputation among people who can tell — has been replaced by algorithmic visibility. The algorithm doesn't grade reliability; it grades engagement. Marius cites a 2024 analysis of >6,500 U.S. state legislators showing that distributing low-credibility information correlated positively with platform attention. The same dynamic works on anyone who needs an audience for work. The careful professional posting once a week loses to the carnival barker posting five times a day, and the gap compounds.

LinkedIn is the cleanest expression. There's now a recognised "mentorship influencer" sub-genre selling hollow advice and bogus referrals to desperate job seekers, all backed by a curated photo and a few thousand bot-inflated followers. The platform can't distinguish a five-paragraph essay by a decade-experienced engineer from a five-paragraph LLM completion sprinkled with emojis; the algorithm sees both as content, and the cheaper, more emotional one wins. The same shape repeats across Medium, X, Instagram, YouTube, TikTok, Substack — different platforms, identical physics.

LLMs as the shipping container

What changed in the last two years isn't the existence of grift but its marginal cost. Marius's framing: LLMs did for bullshit what the shipping container did for global trade — didn't invent it, industrialised it. Anyone with a browser can now produce a thousand confident, syntactically clean words on any topic in under a minute, ship a book to Amazon, an article to a content farm, a thread to LinkedIn, a video to YouTube — and never know what they're talking about. The output passes the only test the distribution channels apply, sounds about right.

This is the supply-side mirror of i-am-an-ai-hater / i-will-never-use-ai / anti-llm-discourse — those pieces argue against the consumption of LLM-generated work as an individual practice; Marius is naming the market structure that makes that work pay.

Bullshit jobs (qualified)

Marius brings in Graeber's 2013 Bullshit Jobs essay but is careful about the strong claims — a 2022 study found <8% of European workers felt their job was useless, well below Graeber's implied 20–60%, and toxic management explained the unhappiness better than pointlessness. The piece of Graeber that survives, Marius argues, is the observation that a large share of professional output now exists to be seen by other people producing output. Slide decks for slide decks, strategy documents about strategy documents, posts about posting. Useless to anyone outside the loop, fully ordinary work to the people in it. That's the "mature form" of bullshittery — not individual lies, a steady-state ecosystem.

What it costs

The part Marius is angriest about is what this does to people who refuse to play. The engineer who insists on shipping things that work, the writer who insists on knowing the subject before publishing, the designer who tests on real humans — all competing in a market tilted against them. The colleague willing to fake the demo and declare victory on LinkedIn looks more successful. Gets the speaking slots, the promotions, the funding rounds. Possibly the Forbes 30 under 30 (which Marius links to a "30 under 30 pipeline to prison" piece, drily).

The cynicism, exhaustion, and quiet bitterness creeping into professional life over the last few years is downstream of this. People haven't stopped wanting to do good work — good work has stopped paying the way it used to. Marius doesn't write off the alternative explanation that bullshit isn't more common, just more observable (one feed, all of it, all at once), but he doesn't think observability accounts for the whole thing.

What he proposes

Marius doesn't pretend he has the answer. Two halves:

• As viewers — pay for the human-made version when you can, and signal substance out loud when you see it. The system starves careful work of the engagement signal it overpays bullshitters with; correcting that is something individuals can actually do.
• As creators — refuse to perform what you don't believe. Don't write the post you don't mean, even though saying no costs visibility. Every honest professional who declines to bullshit is a data point against the trend; there need to be more.

The closing Frankfurt observation: the bullshitter is not embarrassable, because they have no relationship to the truth they could betray. The honest person can be embarrassed because they've made a claim they meant. In a market that has stopped penalising shamelessness, being embarrassable is one of the few remaining markers of good faith. Be embarrassable.

Where it fits

Marius's piece is the sociological / labour-market half of the cluster the wiki already tracks:

• ceo-ai-psychosis — executive-level pathology where tokenmaxxing leaderboards measure consumption, not output; same incentive shape, one organisational layer up.
• ai-great-leap-forward — corporate AI mandates as fabricated-metric-driven loss of institutional knowledge. The backyard-furnace framing and Marius's slide-decks-for-slide-decks framing are the same observation from different angles.
• skill-atrophy-supervision-paradox — what gets eroded by the dynamic Marius names. If the market rewards the loudest claim over the truest, the careful-professional career is the one being out-competed out of existence.
• i-am-an-ai-hater / i-will-never-use-ai / anti-llm-discourse — refusal-side pieces. Marius is sympathetic but moves the analysis off the individual and onto the system: don't hate the player, hate the game.
• average-is-all-you-need / llm-as-average-democratizer — the supply-side mechanism in compact form. LLMs make average output cheap; Marius's piece is what happens once the floor of the cheap-output flood hits the market.

There's also a clean line to contributor-poker and i-dont-want-your-prs — both are about what gets lost when the cost of a unit of work collapses and the trust-signal that used to ride on top of it collapses with it. Marius's be embarrassable is the same posture Loris Cro proposes for open-source review: bet on the contributor, not the contents.

The Frankfurt frame is the load-bearing reframe of the piece. Several of the wiki's existing anti-LLM critiques pivot on it implicitly — the worry isn't that LLMs lie (they sometimes do) but that they're constitutionally indifferent to whether what they're producing is true. Marius makes that explicit and traces the implication out to the market that now selects for the same property in humans.