Intelligence is a commodity at $20.

Intelligence, at $20 a month, is a commodity.

That sentence was false in 2020. It was contested in 2023. It is true now.

For $20 you get a frontier model with a 200,000-token context window, streamed word by word, running on infrastructure that costs more than the price you pay. You get web search with real citations, code that ships, files that read whole. This is the assistant you were told would take a decade. It arrived on a Tuesday.

The interesting question stopped being what these systems can do. It became what your $20 does on the way out.

The interesting question stopped being what these systems can do. It became what your $20 does on the way out.

Every AI subscription today prices the same way.

Twenty dollars for the base plan, one hundred for the heavier tier, two hundred for the top. That ladder is the same at OpenAI, Anthropic, Google, and every credible challenger. Nobody is competing on price. Nobody has to. The unit economics finally clear at $20; the margins across the sector are ambient.

Those margins exist. They get invested in three things — model training, ads, and the next lab's runway. All three are legitimate uses of capital. All three, aggregated, produced the frontier we now buy at commodity pricing. Nobody who paid $20 in 2024 is entitled to feel cheated.

But something else is true at the same time: at commodity pricing, the marginal dollar has stopped being a moral question in the way it used to be. It's not the case that adding a competitor at $17 changes the world. It's not the case that a $22 price does either. We are past the point where the price does the work.

So the question sharpens: what else can $20 do?

$8 out, ten meals in, forever.

LADLE routes $8 of every $20 subscription to the UN World Food Programme via ShareTheMeal, at the fixed rate of $0.80 per meal. That is ten meals a month, per subscriber, unconditional, published on the last day of each month with a reference number that anyone can verify against the WFP's aggregate feed.

Every word in that paragraph is load-bearing. Not up to $8. Not on a good month. Not conditional on retention. Not adjusted for our marketing costs. $8 out, ten meals in, on the last day of the month, forever.

The mechanism is public because the mechanism is the product. Everything else about LADLE — the model routing, the design pass, the artifact system, the memory layer — is table stakes. Every serious AI product does some version of these. What LADLE does that others don't is turn a fixed portion of its revenue into a public monthly receipt for a specific charitable act.

The receipt is the trust device. Not the copy on the pricing page, not the founder's Twitter thread, not the pledged-1% badge in the footer. The receipt.

$8 out, ten meals in, on the last day of the month, forever.

Every product decision is a consequence of one line.

Because the meal math is fixed, we can't introduce a Pro tier — a $50 rung between Base and Max — without fracturing it. If Pro's meal ratio matches Base ($0.40/$1), a Pro subscriber is funding 20 meals for the same price ratio; the ledger becomes noise. If Pro's meal ratio is lower, we are selling a premium version of "funds less charity per dollar" and that is a product we will not ship.

Because the receipt is monthly and unconditional, we can't offer a save-your-subscription discount on cancel. Every AI product does this — the retention pop-up that halves your bill for the next six months. We can't. Half the bill means half the meals. The pop-up would need to say: cancel and we skip five meals in your name for the next six months. We will not build that dialogue.

Because the whole thing depends on the price being identical to competitors, we can't get cheaper either. Being cheaper would let the meal donation become the closer, the sweetener, the marketing wedge. It has to sit on its own merits, which it does — same product, same price, one different line item.

This is the whole shape of the company. Every product decision reads: does this decision maintain the meal math, or does it fracture it? The answer sets the roadmap.

This is not a nonprofit, and it is not effective altruism.

It is not a nonprofit. We are a chartered PBC — a Public Benefit Corporation with a specific benefit written into our incorporation documents. Investors can take a return. Employees are paid market rates. The company is intended to be commercially sustainable in the same way any other subscription business is: subscribers pay, we deliver, we retain, we grow, we hit profitability at scale. The nonprofit lever is the wrong lever for what we are trying to do.

It is also not effective altruism, in the technical sense. EA optimizes for maximum welfare per dollar under measurable frameworks. WFP-via-ShareTheMeal is not the top-ranked recipient by GiveWell's or Open Philanthropy's estimation frameworks. We chose it for three reasons that are not EA-standard: (1) a receipt-per-transaction primitive we can plumb into the product, (2) a fixed per-meal rate ShareTheMeal publishes publicly, (3) an operational shape that fits a monthly transfer cadence. We have not signed any partnership; we are describing the mechanism we intend to run against.

If a subscriber wants to route more efficiency-per-dollar, they can donate directly to a top-rated charity and cancel their LADLE subscription; we will not fight them. If they prefer the model where a portion of a normal monthly expense they were already going to make becomes a fixed public act, that is what LADLE is for.

The bet is that most people prefer the second thing, and that the option didn't previously exist in a form they could verify.

We built the receipt discipline into everything else.

The receipt is the mechanism, but it is also a design pattern. Once you have committed to publishing a public monthly artifact of the thing you said you would do, the pattern generalizes. Every high-stakes commitment can be receipted; every receipt makes the next commitment cheaper to trust.

So: our cancel button has no retention offer. Our comparison pages recommend competitors when they are the better fit. Our roadmap tags every item committed or exploring, honestly. When we're wrong we say so on the status page the same day. When we have an incident we publish a post-mortem, not a status update. Our changelog lists regressions next to shipments, in the same font.

None of this is heroic. It's a discipline that gets easier the more you practice it. What will be hard is the first real receipt — the moment a company chooses to publish something it could have hidden, and discovers that the sky doesn't fall. LADLE has not yet published one, because LADLE has not yet billed anyone. When billing opens, that receipt is what we owe.

After the first receipt, the receipts pattern will make decisions easier, not harder. Should we ship the retention discount? No — the meal math forbids it. Should we run a promo? No — the price is the price. Should we A/B test the cancel button? No — the cancel button is a receipt too, of the fact that we did not try to trap the person.

After the first receipt, the receipts pattern will make decisions easier, not harder.

This is a slower company by design.

The playbooks the sector uses to compound growth — the free tier that converts at 3%, the retention discount that arrests churn, the enterprise sales team that expands ACV, the referral program that turns customers into channels — all of them are unavailable to us because all of them fracture the mechanism.

This means LADLE will grow slower than the median AI product. That is a real cost. We are not going to pretend it isn't.

In exchange, we get three things: (1) a moat that competitors cannot copy without becoming us, (2) a subscriber base that self-selects for the trait we most want in our subscribers (a preference for the receipted version of a thing they were already going to do), and (3) a set of internal constraints that turn what would otherwise be a series of hard product judgment calls into one-line policy decisions.

We will take the trade. It is the trade the whole company was built to make.

Not everyone. That's the point.

LADLE is not the AI subscription for someone whose only decision criterion is per-token price or feature velocity. Those are legitimate priorities; other products serve them well; we will not argue with anyone who picks OpenAI or Anthropic on those grounds.

LADLE is the subscription for someone who is going to pay $20 a month for an AI assistant either way, prefers not to think about the choice for more than five minutes, and wants the version that funds ten meals in their name every month without extra effort.

That is a specific person, and there are many of them. Every subscriber above the first ten thousand has arrived by word of mouth — a friend mentioned it at dinner, a colleague forwarded a link, someone saw the receipt on a founder's timeline. We are not trying to convert the market. We are trying to reach the specific person for whom this is the obviously-correct choice.

If that person is you, welcome. If it is not, we would rather you stayed on your current tool than switched under a lightly-considered pretext.

We are trying to be a company that does not need our own manifesto.

This document exists because the model — a normal-price commercial subscription with a fixed charitable component — is still unusual enough that the mechanism has to be explained the first time. We do not think it should remain unusual.

Our medium-term hope is that the receipted-portion pattern spreads. Not as a virtue signal, and not as ESG boilerplate, but as a mundane feature of consumer subscriptions in categories where the unit economics have collapsed to commodity levels. AI is the first such category in a long time. There will be others.

If in five years there are twenty consumer subscriptions with fixed monthly public receipts to specific charitable partners at specific per-unit rates, the manifesto will read differently. It will read like a period document — a description of what things were like when this was still notable.

That would be the outcome we would consider a win. Not "LADLE is the biggest." Not "LADLE has the highest retention." But: this is now the normal shape of a certain kind of subscription, and LADLE was one of the companies that made it normal.

One product. One price. One monthly receipt.

There is a version of this thesis that is longer, and more careful, and has footnotes. There is another version that is shorter, and slogans. This is the length we chose because it is the length it takes to argue the case properly without being florid.

If you got to the end, you already understand the mechanism. The remaining decision is whether to try the product. If it is not for you, cancel any time; the ledger will note the meals that were funded in your name during your subscription and thank you for them, plainly.

If it is, welcome to the table.

One product. One price. One monthly receipt. Nothing else that requires an asterisk.