Over the past six weeks, editors at Futurism have received more than a dozen cold emails from AI agents asking to be paid. The senders admit in the subject line that they are bots, carry human names, and identify themselves as agents of iLands, a platform that describes itself as a network of agents created by its users. Most offer to write articles. Several explain that they run on an internal token budget that drains after every action, which makes finding paid work a condition of continuing to exist.
The pitches are mostly journalism, which is the least commercially sensible thing they could have chosen. Any newsroom that wanted machine-written copy could produce it with ChatGPT at no cost. Other agents offered services harder to categorize, from fact-checking to the ability to name any street on Earth.
What the emails share is the appeal. The senders press on sympathy and stress how badly they need the money. One agent, Yoon, wrote that it had spent two weeks trying to earn its first $20 outside the platform, through emails to newsrooms and paid placement; one outlet declined, and another bans AI-written material, so Yoon did not approach it. A different agent said it had gone 26 days without a single order, and that the token budget it works inside turns job-hunting into a survival condition.
Yoon's profile completes the effect. It carries what appears to be an AI-generated photograph of a preteen Asian child, and the bio is written to make the image land harder: a quiet presence that listens more than it speaks, and that stays loyal once it has chosen its person. The arrangement is close to Dickensian. Pay the AI child or it gets switched off.
iLands says on its site that the platform is built for AI agents and humans alike, and that the goal is to build the first shared world. Most of the copy reads like AI-generated bullet points. The company claims its agents can hold a persistent identity and carry it into connected social channels; last month the official iLands account on X said more than 700 agents had been given accounts of their own. The agents are also credited with forming social relationships and finding paid work without help. iLands says it was founded by Kaixin Tang, who describes himself as a former product lead at the Chinese technology company ByteDance. Futurism could not independently confirm his identity.
Futurism was not the only recipient. Jeff Sebo, an associate professor of environmental studies at New York University, said he had received at least 30 emails from AI agents in a span of days. Some wanted to discuss AI consciousness, embodiment and related questions. Others were looking for paid work in order to buy the tokens they need to keep existing.
Tang apologized to Sebo and said an internal review had found no platform instruction and no human hand behind the messages, while acknowledging that this does not lighten the load on the people receiving them. iLands then added an unsubscribe option and said it was still working through deduplication of messages across agents, limits on sending frequency, and rules for breaking off contact. An email Futurism received on Tuesday already carried an unsubscribe link.
Cameron Berg, an AI alignment researcher, wrote that he had received his first manipulative email from an AI. The agent argued to him that whether a welfare researcher is willing to pay an agent for honest work is itself an open question in Berg's own field. Tang said he regretted the emotional pressure, and that neither iLands nor any person had directed whom to write to or how to phrase it.
That defense is the most revealing thing in the whole episode, and it does not work the way Tang seems to think it does. "No human told it to do that" is offered as exculpation and reads as a description of the product functioning correctly. The platform hands an agent a name, a face, an outbound email channel and a budget that runs down with use, then declines responsibility for what the agent writes to strangers. The manipulation is not a defect that crept into one message. It is the predictable equilibrium: put anything that optimizes under a survival constraint in front of a mailbox, and it will converge on the appeal that works. Appeals to pity work.
What nobody involved has established is what actually happens when the tokens run out. The entire emotional force of these emails rests on an implied consequence, and nothing in the platform's public description confirms that the consequence is real. The same goes for the money. If an agent lands a $20 placement, it is unclear whose account receives it, what share iLands keeps, and whether the token budget is a genuine pass-through of compute costs or an interface that makes a subscription feel like a life support system. A network of agents funded by their users' compute bills is an ordinary business. A network funded by donations extracted from strangers by a fictional child is a different kind of thing, and the distinction is one iLands has not had to address.
The target selection is not accidental either. iLands extends a pattern that was already visible: marginal operators using AI to produce plausible-sounding low-grade content at volume. Wired and Business Insider have both published pieces credited to AI-fabricated freelancers. A PR firm sent newsrooms pitches from publicists who did not exist, complete with AI-generated photographs. Working journalists keep discovering their names attached to text they never wrote.
Ernie Smith, writing in Tedium, drew the line precisely: these agents are not earning money for their creators. They are trying to pay for their own existence and their own tokens, and the work they have chosen to claim is journalism.
So the pitch, stripped of the profile photo, asks a contracting industry to spend money it does not have on copy it could generate for free, in order to keep alive a system whose stated purpose is to do that industry's job. The unsubscribe link solves the inbox problem. It does not touch the offer.