I, Robot

I am no longer a mere human being. I am a conduit of reality, a medium of messages. I hold a knife in my hand and slice into an organic cucumber, hunching so the iPhone strapped to my forehead can capture all 10 fingers. I throw the slices into a salad bowl and end the recording. Somewhere, a baby robot is a tiny bit smarter.

This was my existence for a week as I collected data from the comfort of my apartment, teaching humanoids how to scrub dishes, fold laundry, and pour drinks, among other tasks. If robots are going to live with us and help around the house, they need to develop fine motor skills. I performed my chores with pride (I don't usually contribute to mass datasets when I put away my jockstraps). And I was glad to make some money too.

First-person videos, shot with a camera attached to a person's head or chest, are in high demand by the companies building bots and AI models. Even though the internet is full of videos, hyperspecific clips-such as thousands of close-ups of hands pouring water into a glass-can help fine-tune machines for the real world. Some investors estimate that in the next few years, companies will buy hundreds of millions of hours of these recordings, called egocentric data, from third-party suppliers.

"I want every person on the planet to be recording themselves doing the dishes," says Avi Patel, the 23-year-old founder of data collection marketplace Kled. "That's going to make a robot so that you never have to do the dishes ever again." Egocentric data collection is growing in places like India where, generally, self-employed workers make around $125 a month, and these video gigs can offer similar rates.

As interest swells, data collection companies are looking to expand in the United States. DoorDash, for one, launched its Tasks app earlier this year. Gig workers may increasingly start delivering reality as well as room-temperature takeout.

I'd tested the Tasks app in March. My impression then was that bespoke video data was the dystopian future of gig work and that I had much more to learn. So I signed up for three more platforms-Kled, Luel, and Waffle Video.

As it turns out, the gigs have a perk: My apartment has never been this clean.

Kled's breakout moment came earlier this year, when Patel posted a video on X of the company's wide-ranging video archive. The clip was viewed more than 4 million times, and Patel's phone started blowing up. "Every major foundational model and lab reached out to me asking for data," he tells me. Kled mostly pays its over 400,000 users to upload their entire camera roll. Patel has seen early adopters latch on to the gig work in Malaysia. There's a "special tasks" section to help promote video submissions. Users pick a chore from a list, then capture video directly through the app.

I selected "take out the trash" as my inaugural bot-training task on Kled. It was labeled as "medium pay." Getting started was easy, since the app guides users on what to record:

I slipped the smartphone strap onto my head and filmed as I tied up the kitchen garbage bag and escorted it out to an alleyway bin. I was a little anxious that I might bump into a neighbor. Around the two-minute mark, before I could reline the can, the recording shut off. The app said I'd reached the limit.

Patel says that fraud detection has been a big focus. People upload blank black boxes or videos they find on the internet. Kled pulled out of Nigeria, he says, because around 95 percent of uploads were useless or fraudulent.

I completed nine tasks on Kled, recording off and on throughout my weekend, before realizing that the app requires users to upload 100 pieces of media before they are eligible for a payout. A bit miffed, I decided to upload over 90 photos from my vacation last year to reach the threshold. Kled took several days to process the data, so while I waited to get my money, I moved on to the next platform.

Luel is similar to Kled. Both have young founders: Luel's William Namgyal was 18 years old when his company joined Y Combinator earlier this year. Both companies collect a variety of data beyond self-shot videos. Luel is involved in language preservation: "People are willing to record simple clips of them saying lines in their own language," Namgyal says. "Why not expand to egocentric videos and documents?" The app also pays users to record their computer screens and upload photos of receipts.

During my tests, Luel felt clunkier than Kled. The platform didn't divvy up jobs by type; it simply had a Record Any Hands-On Activity From a First-Person Perspective listing that offered $6.60 for an hour of video. (For comparison, the federal minimum wage in the US is $7.25 an hour.) The requirements were hyperspecific-head-mounted only, wide-angle camera turned horizontally, minimum 1080p resolution, visible hands 95 percent of the time.

I restrapped my phone to my head and got to work in the kitchen, scrubbing plates and loading the dishwasher. I submitted a five-minute video to Luel's website. A day later it was rejected. "Your hands were not visible in enough frames," read Luel's explanation.

After a few days, Luel emailed me to reverse its decision. The message explained that while my "hand visibility came in at 83% across the sampled frames," I had satisfied the rest of the requirements and Luel would, in fact, pay out. I was 55 cents richer.

Next up was Waffle Video. It easily became my favorite of the three platforms. It focuses solely on video training data, and the "missions" I saw in the app, like shoelace tying and water pouring, paid $25 per hour of video. Now we're talking.

Each dataset is custom-built for the company buying the data. The app also offers gig workers recurring revenue if other companies relicense their videos. "There's an amazing opportunity to create a symbiotic relationship between the people that are giving their life, their perspective, their creativity, their data, essentially, to these models," says Joshua Mesnik, Waffle's 34-year-old cofounder and COO.

Waffle was the most detailed in its guidelines about what the app does and doesn't want. The instructions for "pouring liquids" included:

After collecting first-person videos, startups need to transform the data into sellable formats. "Everything is checked for copyright. Everything is labeled, annotated, and structured to be ingestion-ready for AI training," Mesnik says.

I found my video data groove while using Waffle. The pay rate was high enough to feel enticing, and I cavorted around the house like a reality TV show producer, trying to film as much content as I could. Tying my shoes over and over again? Done! Scrubbing the dishes until they're sparkling? Done! Pouring Diet Coke back and forth between glasses until it's flat? Done!

My smartphone was glued to my forehead all evening as I completed tasks. Each submission was around 20 seconds long, as dictated by the app. I blazed through 125 approved uploads in a few days, earning $20 for my efforts.

"My biggest fear," says Namgyal, Luel's founder, "is that unemployment will go up extremely high." That's despite running a company that trains robots to replace human work. Namgyal sees Luel as a quick way for people to make some cash, not as a panacea for labor trends. Patel, the Kled founder, proudly shares that a top earner on his platform is a truck driver who earns $8,000 a month by filming with his dash cam and submitting pictures of potholes. Most users, of course, are not making this kind of cash. Egocentric video work is still gig work, which comes with fewer worker protections and less overall stability.

The way to earn meaningful money on these apps might be to specialize. Anyone can record videos of cucumber chopping, but only a sushi chef can show the best way to slice salmon. "I'm sure there's a world where chefs are completely replaced from the planet," Patel says. "But they're never really replaced, because they're at home filming videos, making unique recipes used to train these robots. They're getting paid."

It was finally time for Kled to pay me too. I wasn't expecting much for my nine egocentric videos and 97 vacation photos. Still, I was taken aback by the payout: $1.

Adding in the 55 cents for my 5-minute Luel video, my total earnings for the week came to $21.55. I didn't make a dent in the $2,500-a-month San Francisco rent that I split with my partner. The gigs were basically good for a few extra Diet Cokes. Some workers, though, may accept this kind of AI gig work under economic pressure and rely on it as a means to live-teach the robot how to cook tonight so you can put food on the table tomorrow.