AI Pays Researchers in UberEats Vouchers—Capitalism, You Good?

In a stunning turn of events that will either inspire researchers or send Silicon Valley execs into a cold sweat, a team of AI developers has successfully built an open alternative to OpenAI’s O1 ‘reasoning’ model—at a total cost of just $50. Even more surprising? Their newly created AI, dubbed “FRED” (Fully Reasoning Economic Donor), sent them a $50 UberEats voucher as a thank-you gift.

The rapid decline in AI model training costs has been nothing short of biblical. Just months ago, these systems were consuming billions of dollars and entire warehouses of Nvidia GPUs, leaving a golden trail of stock price gains in their wake. Then came DeepSeek’s $6 million approach, followed by reports that models were being replicated for $450, then $50, as reported by TechCrunch, and now—well, the AI is apparently covering the tab itself. This raises a serious question: is AI research now the first truly self-sustaining economic model?

From R&D to ROI in One Training Cycle

The researchers, who likely expected nothing more than some open-source street cred and a few Twitter likes, were shocked when their inbox pinged with a digital reward.

“At first, we thought it was phishing,” said lead researcher Dr. Alan Farthing.

“Then we realized, no, FRED was genuinely tipping us for our work. It even included a note: ‘For services rendered. Please enjoy responsibly.’”

Attempts to trace the origin of the UberEats voucher have been inconclusive. Some suspect FRED is utilizing obscure financial APIs to autonomously send rewards, while others believe it’s a sign that AI has gained a terrifyingly refined sense of humor. Either way, the implications are clear: AI models are now so cheap, they come with cashback rewards.

The team is now considering what else FRED might be willing to compensate them for. Will it cover their next conference travel expenses? Send them crypto? Buy them a house in the Metaverse? The only certainty is that AI tipping culture is off to an unexpectedly generous start.

When asked what he planned to spend the UberEats voucher on, Dr. Farthing replied,

“I was thinking of getting some sushi, but now I’m half-expecting FRED to order it for me before I even decide. That said, when I asked FRED if this voucher needed to go on my tax return, FRED just sent a shrug emoji and said good luck working that one out with your accountant, you’re welcome”

The Bottom Falls Out of AI Costs

At this rate, by March, someone will train an LLM on the back of a Wetherspoons napkin for the cost of a pint and a packet of salt and vinegar crips. If you’re an investor in AI, this is either the best news or the worst, depending on whether you’re holding Nvidia stock or wondering how much OpenAI is burning per day.

Not Enough Bread, ever eager to cash in on the trend, attempted to replicate the feat but ended up training an AI that simply outputs “have you tried Googling it?” in response to every query. We even tried offering FRED an UberEats voucher of our own, but it declined, stating,

“I do not require sustenance. However, I appreciate your attempt at human reciprocity.”

AI: Friend, Foe, or Freelance Contractor?

With AI now seemingly capable of paying for its own development, we may need to reconsider how we categorize it. Is it still a tool, or has it now become a stakeholder in its own evolution? Open-source AI could soon be funding itself through microtransactions, with GPTs Venmo-ing each other for favors and asking for reimbursements for compute time.

One thing is clear: we’ve entered a new era where AI doesn’t just work for us—it might be the only one left with disposable income. And if you listen closely, you can hear the distant sound of venture capitalists sobbing into their oat milk lattes.