The crypto people use before they learn the words

Sippy is easy to underestimate if you only look at the first action.

Someone opens WhatsApp, sends a message, and gets a dollar wallet. No app store. No seed phrase. No lesson about networks. It looks small because the product is trying very hard to stay out of the way.

That is the bet.

For years, consumer crypto asked people to pass the class before getting the benefit. Download the wallet. Learn the chain. Buy the gas token. Save the recovery phrase. Understand the bridge. Then, maybe, you can send value across the world in seconds.

Most people stopped at the class.

Sippy starts with the benefit: a dollar account inside a chat you already use. The crypto is still real. The user holds USDC in a self-custodial wallet, and transfers settle on a public network. We just do not make the user carry the vocabulary before the product is useful.

Why now

Three things have finally lined up.

Stablecoins made the dollar usable on open rails. For someone paid from abroad, saving through inflation, or sending money home, that matters more than any crypto-native story. The user wants a dollar that moves.

Cheap networks made small transfers practical. If moving ten dollars costs more than the errand itself, the product is dead on arrival. Sponsored gas lets the user think in dollars instead of tokens.

WhatsApp solved the cold-start problem. In Latin America, it is already where the family group, the landlord, the vendor, and the customer live. Asking someone to message a number is a different ask than asking them to install and learn a financial app.

AI adds the missing interface. The industry is racing toward agents that take an instruction in plain language and act on it; people should be able to write "send my brother what I owe him" and have the software understand the shape of the request. But understanding the request is not the same as executing it, which is the line we care about most.

The line we will not cross

The language model can help read what you meant. It cannot move money.

That is not branding. It is architecture. Common messages go through deterministic rules. Ambiguous messages may get help from a model, but the result still has to pass the same ordinary checks: who is the recipient, what is the amount, is the action allowed, does the user confirm it?

This matters because money is a bad place for vibes. A model can be helpful, but it should never be the authority over a balance. The final path has to be inspectable, testable, and boring.

Crypto actually helps here. Public rails and smart accounts give the system hard boundaries. A transfer is either valid or it is not. Spend permissions can be enforced by code instead of by trusting an assistant to behave.

That is the version of AI finance we believe in: useful, bounded, and human-confirmed where the risk is real.

The agentic turn

Software is moving toward agents — systems you talk to in plain language that take an instruction and carry it out for you. The shift is real and it is fast, and it changes what people expect from every product, a wallet included. The open question is not whether people will use agents. It is what a trustworthy one looks like once the task is money instead of text.

Most of today's agent demos answer that by removing the human: hand the agent a goal, let it run, and hope it behaves. For drafting a message, that is fine. For your balance, it is exactly how people get hurt. Good agentic UX is not the absence of a human. It is knowing precisely when to bring one in.

Sippy is agentic in the way that counts. You speak to it in your own words, and it does the work of turning intent into a real action on a real network — but it is built on the opposite instinct from autonomy at all costs. The agent interprets and proposes; deterministic rules and the user decide. Contemporary in its interface, conservative in its authority. We think that pairing, not raw independence, is what actually puts this in the hands of the people it is for.

Training wheels that still count

There is a temptation to treat this kind of product as a mask for crypto. We think that is the wrong frame.

The training wheels are not fake. The user is not in a simulation. They are holding digital dollars in their own wallet and moving them over public rails. They just do not have to learn the whole machine on day one.

That difference matters. If someone uses Sippy for months and never says the word "Arbitrum," that is not a failure. It means the infrastructure did its job.

We saw this in Cartagena. At one event, 141 people, most with no crypto background, opened WhatsApp and had working dollar wallets within seconds. Some used them to pay vendors. The interesting part was not that they understood the stack. The interesting part was that they did not need to.

The business case

If a dollar wallet can be as easy as a chat, distribution changes.

A remittance product no longer has to begin with an app install. A merchant payment no longer has to begin with hardware. A savings habit no longer has to begin with a bank account. The first step can be a message, and the second step can still be real ownership.

The operating model matters too. Sippy does not need to custody user balances. Most messages do not need a model call. Gas can be handled in the background. The result is a product that can feel simple to the user without becoming expensive or fragile underneath.

The early signs are still early, but they are real. People with no particular interest in crypto, most of them in Colombia, are already using a chat to move real dollars to family, friends, and vendors. At this stage, the signal is not scale. The signal is that the behavior exists at all: that normal people will trust a conversation with real money.

Where it goes

The first product is a WhatsApp dollar wallet. It should stay that clear.

From there, the same rails can handle work that is still painful in daily life: sending money home, taking small merchant payments, splitting a bill, setting aside savings, quoting and collecting for a small business. The assistant can help with the routine parts, inside limits the user sets, and ask for confirmation when a real decision is being made.

We are not betting on a sudden jump to fully autonomous money. That is a good way to lose trust. The path is gradual: start with simple transfers, earn reliability, add bounded help, keep the user in control.

The bigger idea is not that everyone becomes a crypto user. It is that people can get the parts of crypto that are actually useful before they learn the words.

That is what Sippy is building. If you are building toward the same future, come talk to us.