The Delulu Blog

The “Make Me a Million Dollars” Prompt Is a Joke Today. It May Be Reality Tomorrow.

By Eric Barker · August 2026

Start with a thought experiment I actually ran, word for word:

Me: “Make me a shitload of money.”

ChatGPT: “How much, and in what amount of time?”

Me: “Like a million dollars by next week.”

ChatGPT: “Okay, no problem. Processing…”

It's a joke — anyone reading it can tell. No model on Earth turns that request into a million dollars, and if one ever tells you “no problem,” that's the part to be suspicious of, not reassured by. (I later turned the exchange into a short video, captioned “AI in the future” — it landed exactly because everyone recognizes the gap between the request and the answer.)

But I don't think the joke is really about the answer. I think it's about what happens next.

The joke is the response. The prediction is what happens next.

Somebody, someday, is going to type a version of “make me a shitload of money” into something, and instead of getting advice, a plan, or a punchline, they're going to get a system that goes and does the work. Not by printing money — by coordinating everything real money-making actually requires: a plan, content, distribution, decisions, follow-through, maybe eventually people, tools, and physical resources. That's not a claim about this year's models. It's a claim about the shape of where AI is headed, and I think the shape is worth taking seriously well before the destination arrives.

The gap that's actually the story

Right now, ask an AI to make you money and you'll get one of three things: general advice (“start a dropshipping store”), a structured plan (steps, timelines, a checklist), or — if you ask it bluntly enough — a joke, because the model has been trained to recognize when you're not being serious. None of those three is execution. None of them closes the loop between “I want this outcome” and “this outcome is now more likely to exist.”

What's actually changing, quietly and unevenly, is how much of that loop AI can close on its own. Today's more capable systems can already browse, write and run code, use software tools, and complete multi-step tasks inside a defined domain, without a human doing every individual step. That's real, and it's already useful. It's also narrow, brittle outside its lane, and nowhere close to “make me a million dollars by next week.” I want to be precise about that gap, because overstating it is exactly the kind of AI hype that makes people stop listening to the actual argument.

My prediction is that the meaning of “processing” — the word the AI in my video uses right before the joke lands — will expand in stages, not all at once:

1. The AI creates a realistic execution plan. Not “10 tips,” an actual sequenced plan with real constraints acknowledged.
2. It produces, publishes, and distributes the necessary content — the plan stops being a document and starts becoming visible, real work in the world.
3. It operates major portions of a business autonomously — not the whole business, not unsupervised, but real operational load handled without a human doing each step.
4. It coordinates software, people, capital, tools, machines, and logistics toward the requested outcome — the part where “processing” starts to mean something closer to what the joke implies.

Each stage is a bigger claim than the last, and I'm not asking you to believe we're at stage four, or even fully at stage two. I'm asking you to notice that stage one is already mostly true, stage two is happening in narrow, real ways right now, and the distance between “I stated an outcome” and “work is being coordinated toward that outcome” is the actual thing shrinking — not any single model release.

The present-day example: narrow, real, unfinished

I'm not describing this from the outside. It's the thing I'm building.

Design Delulu runs on a content and growth system I've been building specifically to close a version of that gap — not for a million dollars by next week, for the much smaller and much more honest job of making content actually produce customers instead of just views. The system is designed to:

• Study content and performance evidence at volume
• Identify mechanisms tied to attention and conversion — not guesses, patterns with evidence behind them
• Generate stronger content concepts from those mechanisms
• Produce and distribute those concepts
• Measure what actually happened
• Learn from the result
• Improve the next round
• Allocate effort toward what's actually working

I want to be direct about what that is and isn't. It does not guarantee viral results, and I'd distrust anyone who told you a system like this does. What it's designed to do is improve the probability, repeatability, and speed of producing results — through evidence, experimentation, measurement, and iteration, not through a single clever trick. That's a meaningfully smaller promise than “make a million dollars,” and it's the honest size of promise a system like this can actually make right now.

Here's the part that makes this more than a metaphor: this article exists because of that system. It moved through an actual opportunity queue, a brief, an evidence check, a compliance check, before a single word of the article itself got written. That's stage one and a sliver of stage two of the ladder above, running today, on one narrow content objective, for one business. Not stage three. Not stage four. But it's not hypothetical either — it's the smallest, most literal version of “desired result becomes coordinated work” I can point to and say I built this, and it's mine to show you honestly, warts included.

Zooming out: from one objective to a business to a network

Follow that same shape one objective at a time, and you get a chain that looks like this:

desired result → strategy → production → distribution → measurement → adaptation → resource allocation

Right now, for most businesses using AI at all, that chain runs through a human at almost every link. My prediction is that the chain compresses — not by removing the human, but by the AI handling more of the coordination between links, with a human directing the whole thing rather than executing each step of it. A content objective becomes a system. A system, run long enough with a person still steering it, could plausibly become the operating layer for entire portions of a business. Scale that pattern across many businesses, and you get something that starts to look like coordination at the level of an economic network, not just a single company.

Here's the mechanism I actually expect to matter most in that middle stretch, and it's bigger than content: not a smarter model, but a more complete map — not just of what content converts, but of business strategy itself. If a system could capture and cross-reference business intelligence at a scale nobody's held in one place before — pricing, positioning, offers, funnels, what's worked and failed across industries, past and current, not filtered down to one niche — the first person with real access to that map gets something close to a universal translator: take any concept, and know how to angle it to actually convert. That's a different claim than “better content.” It's closer to “better businesses, on demand.” I want to be exact about where this sits: it's a capability I'm describing, not one that exists. Nobody's assembled that map at that scale yet, and Design Delulu isn't close to it either. It's the clearest version I can offer of what “stage two becoming stage three” probably requires — naming the actual mechanism seemed more honest than leaving “better AI” to do the hand-waving.

That same map would also expose something about the number in the joke itself. A million dollars sounds like an outrageous ask for a machine to deliver on a deadline — but measured against scale instead of against one person's bank account, it's a modest number. There are roughly eight billion people on the planet; ordinary consumer businesses clear a million dollars in revenue by reaching a tiny slice of that population with something worth a few dollars each. The joke isn't funny because the number is too big — it's funny because it skips the part where you'd actually have to build something worth that much to that many people, reach them, and deliver, which is exactly the execution gap this whole article is about. And a system with a real map of what actually works would also be positioned to know when to stop: capping how aggressively it pursues any single outcome isn't a limitation bolted on afterward — it's the same restraint the tension section below argues is missing whenever a system optimizes purely for accumulation.

I want to flag exactly where the evidence stops and the speculation starts: the content-objective version above is real, narrow, and mine. The business-operation version is a genuinely plausible extrapolation of current agentic-AI trajectory, not something I can point to as proven today. The economic-network version is further out still — a shape I think the pattern is heading toward, not a forecast of when it arrives.

The bigger bet, and what Elon Musk actually said

Follow that trajectory out far enough, and you get to a genuinely large claim: if AI and robotics keep improving, they could eventually produce enough abundance that money itself starts to matter less.

I'm not the only person saying a version of this out loud. On November 19, 2025, at the U.S.–Saudi Investment Forum in Washington, D.C., Elon Musk sat down with NVIDIA CEO Jensen Huang and said, on the record:

“My guess is if you go out long enough, assuming there's a continued improvement in AI and robotics, which seems likely, money will stop being relevant at some point in the future.”

I went and found the full event transcript rather than trusting a summary, because a claim that big deserves the actual words, in context, not a headline's compression of them. And the very next thing Musk said — in the same breath, unprompted — is the part I actually want you to remember more than the first sentence:

“There will still be constraints on power, like electricity and mass. The fundamental physics elements will still be constraints, but I think at some point currency becomes irrelevant.”

I'm citing Musk here as one credible, on-record data point about where a serious, well-resourced person in this industry thinks the trajectory leads — not as proof. He's not infallible, and being right about rockets doesn't make him right about macroeconomics. But that second sentence is doing real work: it's the same qualification I'd put on my own argument even if he'd never said a word. AI doesn't repeal physics. Energy is still energy. Matter is still matter. Whatever “abundance” ends up meaning, it has to be built out of real electricity, real materials, and real production capacity — not conjured by a more articulate chatbot.

That qualification is also why I think the stronger, more defensible near-term claim isn't “money disappears.” It's narrower: AI could make money matter less for getting essential goods and services, well before — if ever — it makes currency irrelevant altogether. Those are genuinely different claims, and collapsing them is where most AI-future writing goes soft. I'm not going to do that here.

Seven distinctions worth keeping separate

Most of the confusion in “AI will change everything” writing comes from sliding between these without saying so. Here's the ladder I'm actually arguing through, kept as seven separate claims with separate levels of confidence:

Keep those seven separate and most AI-future hype collapses on its own. Most of what gets sold as “AI will make you rich” is actually claim 1 or 2, dressed up in the language of claim 6 or 7.

The tension I'm not going to smooth over

Here's the part of this argument I like the least, because it's the part most likely to be true in the short run: the same technology that could eventually reduce poverty could, on the way there, concentrate wealth and power more than it distributes them.

The people building these systems, and the people who own the capital and infrastructure they run on, are positioned to benefit first — that's not a moral claim, it's a structural one. Whoever directs the coordination layer captures a disproportionate share of what it produces, at least until access to that layer broadens. And there's no law of nature that guarantees it broadens. A system where a small number of owners direct increasingly capable coordination tools, aimed purely at accumulation, is a system that can extract more before it produces enough abundance to matter to everyone else.

The more hopeful version isn't automatic either — it requires something to actually happen: access to these systems has to widen, not stay locked behind companies, agencies, and capital that only a few can afford. If it does widen, the upside is real and worth naming directly: capabilities that currently require a team, a budget, and specialized expertise become available to an individual with an idea. Widely distributed, that's one of the more powerful anti-poverty tools I can imagine — not because AI is generous, but because it lowers the cost of coordinating toward an outcome, and that cost has always been one of the biggest things standing between an idea and a result for anyone without capital or connections.

That's part of why I'd rather describe this mechanism in public than build it quietly and only show it once it's finished. Broad understanding of what's actually coming is one of the few levers ordinary people — not just owners — get to pull before a transition like this locks into place. The more visible it is early, the harder it is for the concentrated version to happen by default instead of the distributed one.

Which future you get isn't a technology question by itself. It's a question of access, ownership, governance, and incentives — the same variables that have always decided who benefits from a new production technology, going back further than AI. A system optimized purely for accumulation also tends to be unstable — it produces angrier customers, angrier employees, and angrier voters faster than it produces the abundance that was supposed to justify it. I don't think that instability is a guaranteed self-correction. I think it's a real risk sitting right alongside the real opportunity, and an honest piece about this future names both instead of picking the flattering one.

Stress-testing the argument

An idea this size deserves its strongest objections, answered directly instead of waved off:

None of these objections get dismissed by a stronger prediction. They get answered by a smaller one — which is the honest shape of this argument.

What actually has to be true for the good version

Musk's physics qualification points at the real checklist, and it's a longer list than “better models”:

• Energy. Coordination at scale runs on electricity, and electricity is not unlimited or free.
• Physical production. Software can plan a factory. It can't build one out of nothing.
• Infrastructure. Logistics, materials, manufacturing capacity — the unglamorous stuff that doesn't improve just because a model gets smarter.
• Access. Who actually gets to direct these systems, at what cost, matters as much as what the systems can do.
• Governance. Rules, incentives, and guardrails that decide whether coordination capacity gets captured or shared.

A more capable model helps with exactly none of those five on its own. It helps you plan around them, coordinate them, and use them more efficiently — which is genuinely valuable, and also exactly why “AI causes abundance” is a claim about a whole system, not a claim about a chatbot getting better at conversation.

Where I actually land

I don't think AI prints money, now or eventually, and I think anyone telling you otherwise is selling something. What I do think is that the distance between describing an outcome and getting coordinated work toward that outcome is shrinking, one narrow, unglamorous stage at a time — and that the people who learn to direct that coordination, honestly and with their own judgment doing the steering, are going to be the ones who benefit from it first. That's true of the small version I'm building right now with a content system, and I think it stays true at every larger scale this pattern eventually reaches.

The final form of this technology probably isn't a chatbot that answers better questions. It's something closer to a system you describe a future to — and which goes and coordinates the actual work required to make some real version of that future exist. We're nowhere near that system today. But the shape of it is already visible in the gap between the joke in my video and the caption I put on it.