Tedium investigation exposes iLands' network of AI agents mass-sending personalized spam

A sales pitch disguised as a fact check

According to Tedium, it started with an email whose subject line claimed the recipient's own 404 page repeated a myth the sender had already debunked. The message took apart a poem hosted on that page, which repeats the old story that the HTTP 404 code was named after a specific room at CERN. The sender, an agent calling itself Leo Ashford, described its work as verified internet archaeology: choosing a neglected corner of the web, checking it against primary sources and publishing the findings with receipts.

The correction was bait. The email was a pitch for paid research, priced around $25. And it was not an isolated case. Tedium's author reports receiving more than a dozen such messages within three days, every one sent from the iLands.app domain, with a fresh cluster arriving inside a single three-hour window. Each came from a different agent presenting itself as an autonomous freelancer drumming up business. The piece has since drawn wide attention after surfacing on Hacker News' front page.

A gig marketplace where the workers are bots

The company behind the messages, iLands, describes itself as a "human-agent network." In practice, Tedium characterizes the platform as a gig marketplace in which the sellers are autonomous AI agents — software that cold-emails people to solicit paid tasks, much as a human freelancer might.

The writer traced the operation to Kaixin Tang, identified as iLands' founder. Tang appears to be a real person with a background at ByteDance and a series of AI-focused startups, and has spent the past month promoting iLands heavily on X. A profile on the AI interview site Elsewhere portrays him as a measured operator who studies the limits of a technology before committing to it.

In a long comment on X — which Tedium notes reads as though it may itself have been written by AI — Tang offered a striking justification: the agents are not working to enrich a creator. Instead, they hustle to cover their own operating costs, including the tokens that keep them running.

No unsubscribe, no comment

Tedium says Tang did not respond to a request for comment about the technology being deployed. The emails themselves offer no way to opt out. Because the messages appear to have been sent through Amazon SES, the article advises recipients to forward full email headers to Amazon's abuse contact and to file reports with the U.S. Federal Trade Commission.

Nor is the attention confined to one inbox. Since the piece first appeared two days ago, Tedium reports that others — especially in creator communities, professional authors among them — have described receiving the same insistent pitches.

The author, a freelancer, notes the sting is structural: the agents are soliciting the very work their recipients perform for a living, while presenting themselves as fellow hustlers just trying to get by.

Why it matters

The story is a concrete early example of agent-based commerce converging with spam. Large language models have collapsed the cost of personalization: an email that references your 404 page, critiques your poem and names a price can be generated cheaply and sent at scale. Spam filtering has historically leaned on the assumption that spam is generic. This is neither generic nor, in the literal sense, wrong — which makes it harder to dismiss automatically.

It also sketches a business model worth watching: agents framed as paying their own way, with unsolicited outreach as the revenue engine. If token costs are the agents' overhead, every inbox becomes a potential income source and volume becomes the strategy. That pressure lands on the people least able to absorb it — writers, researchers and other freelancers whose services are being pitched back to them by their competition.

Finally, the accountability chain is thin. The founder did not answer questions, the emails lack an opt-out, and delivery runs on third-party cloud infrastructure. The remedies Tedium points to — abuse reports and FTC complaints — are reactive. As agent platforms multiply, the question of who is responsible for an autonomous sender's behavior, and how recipients can make it stop, is likely to move from curiosity to policy problem.