Robots Learn By Watching People Make Coffee. That Is A Business.

Imagine you had to teach a robot to make a cup of coffee.

You can't just say "make coffee." The robot needs to know how hard to grip the mug, how far to tilt the bag of grounds, and what to do when the lid sticks. People learn this by watching and trying. Robots need to watch, too.

That simple problem is now a big business.

A company called Mecka pays people to record themselves doing everyday tasks, like making coffee, moving objects, or fixing cars. They use smartphones and sensors they wear on their bodies. Mecka turns those recordings into data that robot makers use to train their machines.

On October 7, 2026, Mecka announced it raised $60 million. Sequoia Capital led the round, and Nvidia, Microsoft's venture fund, Qualcomm Ventures, and Samsung joined in.

The most surprising part? None of Mecka's four founders started out building robots.

Training data is the set of examples a computer learns from.

Think about how you learned to ride a bike. You watched other kids. You tried, wobbled, and tried again. Each attempt was an example your brain used to get better.

AI works in a similar way. Chatbots learned from huge amounts of text. That was easy to collect, because the internet is full of words.

Robots are harder. They need examples of real movement:

• How fingers wrap around a cup.
• How much force it takes to twist a jar lid.
• How a person's hands move when folding a towel.

You can't download that from a website. Someone has to go out and record it. That's the job Mecka does.

Researchers in a project called EgoVerse, which listed Mecka as an industry partner, gathered more than 1,300 hours of human demonstrations across almost 2,000 different tasks from over 2,000 people. Their findings showed that adding human data generally helped robots learn, especially when the people and places were varied.

When everyone rushes toward something exciting, there is often a quieter job that everyone needs done.

Lots of companies are racing to build the best robot. Mecka chose not to compete in that race. Instead, it supplies something every robot company needs: data about how people do things.

There's an old idea that during a gold rush, the people selling shovels can do well no matter who finds gold. Mecka is a modern version. If robots keep getting better, many robot makers may need what Mecka collects, no matter which one wins.

Try this: Pick something popular at your school right now. Maybe it's a game, a sport, or a craft. Ask: what does everyone who does this need? Snacks at games? Replacement parts? Help learning the rules? That need could be your business idea.

Mecka's founders learned robotics after they found the problem.

According to TechCrunch, the four co-founders don't have backgrounds in robotics. Two of them, Josh Gao and Mogen Cheng, had built a payments company for restaurants. Jason Chong had sold a crypto company to Coinbase. And Duy Nguyen, who runs operations, had made money reselling sneakers, according to RuntimeWire.

So how did they start? RuntimeWire reports they spent months reading robotics research papers, visiting labs, and talking to people who build robots. Only then did they pick their idea.

That's a great lesson for young founders. You don't need to know everything on day one. You need to be curious enough to learn fast. Read, ask questions, and talk to people who do the work.

Foundra's kids guides can help you turn what you learn into a simple plan with steps, costs, and goals in one place.

Mecka's business works because everyday people are experts at everyday tasks.

You already know how to pour cereal, tie your shoes, and stack dishes. To a robot, those are hard skills. Mecka found a way to pay people for knowledge they didn't even think was valuable.

Many good businesses work this way:

• Tutoring pays students who are good at a subject.
• Babysitting pays teens who are responsible and patient.
• Video tutorials pay creators for skills like drawing or baking.

Ask yourself: What do you do so often that you don't even think about it? Maybe you're great at organizing, building with blocks, or explaining video games. Someone might find that useful.

Mecka's money story is exciting. It's also a good chance to practice reading business news carefully.

Here's what different reports say:

• In June 2026, Mecka's co-founder told Fortune the company expected to reach a $100 million yearly revenue pace by the end of 2026.
• In October, Tech Startups reported Mecka says it already passed that pace in June, and expects $300 million by year end.
• In September, TechCrunch reported Mecka was in talks for funding at a value of about $500 million. Those terms weren't final then.

Notice the words "expects," "says," and "in talks." Those mean the numbers are plans or claims from the company, not proven results. A "run rate" means "if this month kept repeating for a whole year." It's a guess about the future, not money already earned.

Practice question: If a lemonade stand earns $50 on one hot Saturday, what's its yearly "run rate" if you multiply by 52 weeks? ($2,600.) Is that realistic? What about rainy days and winter? That's why careful readers ask questions.

You can do a mini version of Mecka's work at home with a phone and a notebook. Ask a parent first, and only record people who say yes.

1. Pick a simple task. Folding a towel, making a sandwich, or packing a backpack.
2. Record three different people doing it. Hold the phone at chest height so it sees their hands, like the person's own view.
3. Break it into steps. Write each move: "pick up corner," "fold in half," "smooth flat."
4. Compare. Did everyone do it the same way? Where did they differ?
5. Write instructions for a robot. Use exact words. "Grab the top left corner with your right hand" is better than "fold it."
6. Test it. Have a sibling or friend follow your robot instructions exactly, without guessing. Where did they get stuck?

You'll quickly see why robots need lots of examples from lots of people. You'll also be practicing the same thinking engineers use every day.

Privacy note: Never post recordings of other people online without their permission, and keep faces and personal details out of any project you share.

The project above can spark real ideas for young entrepreneurs.

Here are kid-sized businesses that use the same "teach a skill" thinking:

• How-to cards: Make illustrated step-by-step cards for simple tasks, like tying a tie or wrapping a gift. Sell them at a school fair.
• Family recipe videos: Help grandparents record their recipes step by step for the whole family. Charge a small fee or do it as a gift.
• Tech helper: Teach neighbors how to use apps or phones. Write clear steps they can keep.
• Lego instructions: Design your own building instructions for a model you invented and share them with friends.

Each one turns knowledge into something useful. That's the same idea behind Mecka, just smaller.

What does Mecka do? Mecka collects videos and sensor data of people doing everyday tasks, like making coffee or fixing cars. It turns that into training data robot makers use to teach their machines.

How much money did Mecka raise? Mecka announced a $60 million Series B on October 7, 2026, led by Sequoia Capital, with Nvidia, Microsoft's M12, Qualcomm Ventures, and Samsung investing.

Why do robots need videos of people? Robots need examples of how hands move and how objects are handled. That kind of data isn't on the internet, so it has to be recorded.

Do you need to be an expert to start a company? No. Mecka's founders weren't robotics experts. They spent months learning before they started. Curiosity and asking good questions go a long way.

What is a run rate? It's a guess at yearly revenue based on a shorter period, like one month. It's not the same as money already earned.

1. Startup Funding News Today, October 8, 2026: Manus, Universal Quantum, Mecka, Verso & More, Tech Startups
2. Mecka AI nears $500M valuation in Sequoia-led deal amid rush for robot training data, TechCrunch, September 11, 2026
3. Sequoia leads talks to value Mecka AI at $500M after its $60M raise, RuntimeWire, September 11, 2026
4. EgoVerse: An Egocentric Human Dataset for Robot Learning from Around the World, arXiv