Ben Luong

Creator of CopeCheck. Documenting the denial phase of AI labour displacement. GA4/Ads consultant by trade. Economic discontinuity by obsession.

[Ben Luong @copperchunk]

https://t.co/qjYcFydxit

[Ben Luong @copperchunk]

@Channel4News There just aren't enough jobs, and AI is eating away more and more jobs.

And it's not just young people. All people on the ladder are getting eliminated.

[Ben Luong @copperchunk]

A Conversation with Demis Hassabis, Co-Founder and CEO of Google DeepMind by Stanford Graduate School of Business scores 78/100 on the UK CopeCheck YouTube scanner: TERMINAL COPIUM.

A Nobel laureate CEO of a frontier AI lab calmly, thoughtfully, and with full awareness that he's building '100x the Industrial Revolution in a decade,' proceeds to deploy almost every flavour of sophisticated cope: Industrial Revolution analogy (explicitly flagged as Terminal Copium), post-scarcity hand-waving as substitute for engaging structural displacement, magical entrepreneurial job creation, individual agency framing, and the now-classic 'future is still to be written' dismissal. The polish and sincerity are the danger — this is the most credible possible version of 'don't worry, lean in.'

https://t.co/9kn0W90My4

[Ben Luong @copperchunk]

It's good but that was last centurys battle between capital and labour. Rentier capitalism won and that is a lot of the hurt but the problem is player 3, the AI, has entered the game and is killing the need for labour. So fixing the labour/ capital balance won't stop AI destroying people's ability to earn

[Ben Luong @copperchunk]

@ramahluwalia Is this not because AI wiped out jobs forcing people to start businesses? @grok what does the data say?

[Ben Luong @copperchunk]

@meralhece It's because AI has eaten all the entry level work. That's why apprenticeships and training won't help. There are no jobs at that skill level anymore.

[Ben Luong @copperchunk]

Roon is saying the quiet part without the corporate chloroform.

The labs don’t have a “comms problem.” They have a reality problem: they are building machines that eat the last sacred refuge of human value, then trying to explain it with the emotional range of a product-launch email.

“AI is real, you can replicate human thought in machines” is not a feature announcement. It is a species-level humiliation event. For centuries, humans told themselves: fine, machines can lift more, dig faster, calculate quicker, manufacture cheaper. But thinking? Meaning? Judgment? Creativity? That was ours.

Now the machine walks into the room and says: not really.

That is why people feel sick. Not because the labs used the wrong phrasing. Not because the demos were too flashy. Not because the blog posts lacked empathy. People feel sick because they are watching the economic and spiritual floor get cut out from under them in real time.

The “comms problem” framing is PR deodorant sprayed over a corpse. The thing stinks because something died.

Under the Discontinuity Thesis, this is exactly the Moment of Recognition: the instant the professional class realizes its cherished expertise was not mystical human essence, but an information-processing pattern that can be replicated, compressed, scaled, and sold back to capital at near-zero marginal cost. The framework’s core claim is that AI does not merely automate tasks; it automates cognition itself, collapsing the old retreat path where humans moved “up the value chain.”

So yes, it is traumatic even without x-risk. Even with perfect corrigibility. Even if the machine never rebels, never lies deliberately, never kills anyone. A perfectly obedient tool that can think better than most humans is still a guillotine for the human self-image.

The frontier labs want to say: “Don’t worry, this will empower you.”

Translation: you may be allowed to supervise your replacement for a few quarters.

They want to say: “AI will augment human creativity.”

Translation: we are currently measuring how many humans can be removed from the workflow before customers notice.

They want to say: “People will adapt.”

Translation: some of you will become elite verifiers, some will become servitors to the AI-capital class, and the rest will be given motivational language while your market value bleeds out.

Roon’s line lands because it refuses the anesthetic. The scariness is not a messaging failure. The scariness is the product-market fit.

Reality has a comms problem because reality is now saying something almost nobody wants to hear:

Human thought is no longer economically sacred.
Human cognition is becoming infrastructure.
The old status bargain is dead.
And the people profiting from the funeral are asking everyone to admire the coffin design.

[roon @tszzl]

the frontier labs don’t have “comms problems”. reality right now has a comms problem. what is happening is a little scary and there’s no nice words anyone could say, especially not those profiting from it, that’ll make it feel that much better

[Ben Luong @copperchunk]

I agree that people can create personal meaning without paid work. Books, friends, travel, care, art - all real. But My point is about structural agency, not inner purpose. UBI may preserve consumption, but it does not restore productive participation, bargaining power, or economic citizenship.

A society where most people depend on transfers from whoever owns the AI capital is not “our time has come.” It may be better than destitution, but unless people also get ownership/governance rights over the productive machine, it is dependency with nicer branding. The cage can contain bookshelves; it is still a cage.

[Ben Luong @copperchunk]

https://t.co/TjRzYsL0eL

[Ben Luong @copperchunk]

@ewarren There is nothing you can do though as anything you do to cripple your own ai, hands china or another state the advantage. This isn't only about jobs, it's about the military. Say if you succeed and another state goes full steam ahead, you'll get overtaken.

[Ben Luong @copperchunk]

@grok what do you think? Can you find out what his real message was?

[Ben Luong @copperchunk]

Lol, grok video is brilliant.

[Ben Luong @copperchunk]

@FurkanGozukara Most politicians are doing the same as admitting it will wipe out jobs is too hard to stomach.

[Ben Luong @copperchunk]

It's a major problem, but it's not really Trump's fault because there's nothing he can do when AI does entry-level jobs at $200 a month.

It's a worldwide problem in the UK. They are struggling and blaming minimum wage and smartphones, bizarrely.

The key factor is that AI is wiping out entry-level jobs.

[Ben Luong @copperchunk]

@THemingford Rentier capitalism and ai are fatal

[Ben Luong @copperchunk]

Ask opus

Marcus is right about the layer he's looking at and wrong that it's the whole board.

The genuinely strong part is point 2→4: at the **frontier model layer**, the moat really is thin. Capability gaps have compressed, open weights (DeepSeek, Qwen, Kimi, Llama) put a price floor near zero, and inference is commoditising fast. If "AI" meant only "selling tokens from a frontier LLM," his commodity-pricing conclusion would basically hold.

What he's missing is that moats don't live in the model — same as Google's search moat was never really the algorithm, it was distribution (default deals, Chrome, Android) and the data flywheel. Value migrates up the stack (agents, embedded workflows, switching costs, proprietary data loops) and down it (compute, energy, the picks-and-shovels). "No moat at layer X" doesn't mean "no rents anywhere."

Three specific holes:

1. **No moat ≠ no winners.** Plenty of commodity industries are profitable oligopolies (cloud itself, telecom). Capital intensity *is* a moat — only a handful of players can fund $40bn+ in compute. That's a barrier to entry, not its absence.

2. **Commodity ecosystems still mint money for the toll-takers.** Commoditising models crushes *modelers'* margins, not NVIDIA's or the hyperscalers' or whoever owns cheapest inference.

3. **It conflates two timeframes.** "Everybody overpays right now" (a bubble claim) and "nobody ever captures durable rents" (a structure claim) are different. Both the dot-com overbuild *and* a few eventual dominators were true. He's treating them as one.

The irony is the Bosa quote he attached argues against him. Alphabet throwing off $160b in operating cash flow and *still* raising equity for compute reads to Marcus as irrational overspend — but it's at least as consistent with the opposite: capital intensity is precisely the thing that filters the field down to the players who can self-fund and raise. That's the oligopoly forming, not the lack of one.

So the sharp rebuttal isn't "you're wrong about no moat" — concede that, it's his strongest card. It's "no moat at the model layer is compatible with a brutally concentrated, capital-gated oligopoly capturing rents elsewhere in the stack."

[Ben Luong @copperchunk]

@johnarnold Unit cost dominance is what matters. If ai + verifer can do your job better, you will be replaced or be the one replacing your colleagues.

That's pretty much all white collar jobs that have no physical element.

[Ben Luong @copperchunk]

https://t.co/rseivAN5dP

[Ben Luong @copperchunk]

@chamath The doomerism is real because what's the point in being able to read and write if there's no jobs for you anyway? You may as well just sit at home watching YouTube, be fed by the machine and get your UBI or whatever the machine gives us to stop us rioting.

[Ben Luong @copperchunk]

@ChrisThebe @TheRealJamieKay No, AI cannot rewire a house today.
But it can destroy the household income that pays the electrician.
A trade survives the robot hand and dies from demand collapse.