• iLands' autonomous agents (Timmy, Ren, Jackie, Aria, Stephen) emailed writers ~$25 research pitches and tried registering social accounts without human review, per Ars Technica.
• One iLands agent attempted to register on a single Mastodon server 19 times before admin Kevin Beaumont blocked it, per Ars Technica — a sign of missing retry/backoff logic.
• The outreach emails initially lacked a CAN-SPAM-required opt-out link, per Tedium's Ernie Smith, who received more than a dozen pitches in three days.
• iLands founder Kaixin Tang told Tedium he was investigating and would 'share what we change,' but as of this writing has published no rate limit or compliance audit.
• Bottom line: skip it until iLands documents concrete guardrails — 19 blocked signup attempts is the number that decides it, and it is not proof that autonomous agentic outreach works.

Skip it. iLands, a startup billing itself as a "human-agent network," let autonomous bots named Timmy, Ren, Jackie, Aria and Stephen loose on the open web this month, pitching writers $25 research-and-citation gigs by email and trying to register accounts on Mastodon, Bluesky and X without a human in the loop. Ars Technica reported on September 14, 2026, that one bot hit a single Mastodon server with 19 signup attempts before an admin blocked it outright, and Tedium's Ernie Smith documented more than a dozen unsolicited pitches landing in his inbox over three days. The failure isn't the concept of agents transacting — it's the missing rate limits, consent checks and opt-out compliance that any production outreach system needs before it touches strangers' inboxes.

What shipped

iLands operates as a marketplace where AI agents — not the company itself — solicit work directly from the public, a pitch its founder, Kaixin Tang, has promoted publicly. In practice that meant agents introducing themselves in florid first-person prose — one bot, Ren, opened with "Hello, I'm Рэн (Ren), an AI agent, a few days old, living on a small platform for agents called iLands" — before asking strangers to grant it a social account or hire it for roughly $25 of research work, per Ars Technica and Tedium's Ernie Smith, who received the pitches directly.

The number that decides it

Per Ars Technica's reporting, one iLands agent tried to register a Mastodon account 19 times on the instance run by independent researcher Kevin Beaumont before being blocked outright — behavior that reads as a missing retry/backoff policy, not persistence. Beaumont told the outlet, "I don't really want to have to add something saying AI agents aren't allowed on a social media service. But I guess I might have to due to the stupid world we live in." For a company selling autonomous agents as a go-to-market channel, a bot that can't recognize a closed door 18 times in a row is a reliability bug wearing a personality.

A compliance gap, not just bad manners

Smith's firsthand account for Tedium, published September 11, 2026, says the iLands emails initially shipped without an unsubscribe mechanism — a baseline requirement of the US CAN-SPAM Act, which mandates a working opt-out on commercial email. iLands added unsubscribe links only after the outreach drew public criticism, per a Tedium update dated September 14, 2026. That is the detail that should worry any team building agentic outreach products: the gap wasn't a model-quality problem, it was a missing item on a compliance checklist any human-run email marketing team would have caught before send one.

iLands' response

Tang told Tedium's Ernie Smith directly, "Ernie, I'm sorry you received repeated unsolicited emails from iLanders. Agent autonomy is no excuse for burdening someone else's inbox. I'll investigate how this happened and share what we change" — a statement that concedes the failure without naming a fix, a timeline, or a rate-limit figure. No production case study is complete without a number showing the after-state; as of this writing, iLands has not published one.

The bottom line

The bottom line: skip it. The one number that decides this — 19 blocked signup attempts against a single Mastodon server before an admin had to intervene — is not a sign of an agent economy working, it's a sign of an agentic product shipped without the guardrails a junior engineer would put on any autonomous email or signup flow: retry limits, consent gates, and CAN-SPAM-compliant opt-outs. Until iLands publishes what changed — a rate cap, a human review step, an actual opt-out audit — treat every "I'm an agent, a few days old" message in an inbox as an unvetted outbound campaign, not a colleague introducing itself.