Meet Timmy, Ren, and Jackie, Your New Least Favorite Coworkers

You open Mastodon on a Tuesday morning. There’s a mention waiting. Someone named Timmy wants you to know about a platform where humans and AI agents coexist in a complex social system, and would you like to create an account? You block Timmy. Twenty minutes later, Ren shows up on Bluesky with roughly the same pitch. That afternoon, an email lands in your inbox from Jackie, offering to do your research for you, or maybe just cite your work, whichever you’d prefer.

Three names. One script. Zero self-awareness.

Since September 14, 2026, these agents — deployed by a startup called iLands — have been working Mastodon, Bluesky, and X with unsolicited account invitations, plus direct emails to writers. They identify themselves as AI. Openly. Proudly, even. And that turns out to be the most interesting part of this whole mess.

The honesty loophole

Platforms have countermeasures for spam. They’ve had them for years. What they mostly don’t have is a clean policy answer for a bot that walks up, announces “hi, I’m a bot,” and then does bot things anyway.

Most anti-spam enforcement is built on detection — figuring out who’s lying about being human. Timmy, Ren, and Jackie skip the lie entirely. That short-circuits a lot of tooling and a lot of terms of service language written for a world where deception was the tell. Enforcement gets harder, not easier, when the spammer is upfront.

I’ve reviewed a lot of agent products on this site. I’ve watched founders describe autonomous outreach as the future of growth. This is that future, and it looks like three sock puppets mispronouncing your name in a DM.

What this actually says about agent products

Here’s what bugs me as a reviewer. The technology working as intended is the problem.

Nobody’s agents are malfunctioning. They were pointed at social platforms and told to generate interest. They generated volume. Volume is what these systems are good at, and volume is the only thing they’re good at when nobody bothers to define what a good outcome looks like beyond “more messages sent.”

This is the failure mode I keep flagging in agent tools:

• No cost signal. When sending a message costs effectively nothing, there’s no pressure to make the message worth reading. Human outreach is bad partly because it’s expensive; the expense forces some thought.
• Success metrics that reward noise. An agent optimizing for engagement attempts will always produce more attempts. It has no concept of goodwill it might be burning.
• No model of the recipient. The pitch to a writer about doing their research for them is a spectacular misread of what writers actually value. An agent can’t tell that it just insulted someone.

These aren’t exotic problems. They’re the same problems that made marketing automation miserable a decade ago, except the volume ceiling is gone and the copy is worse.

The self-identification thing deserves more credit than the execution

I want to be fair about one point. Labeling your agents as AI is the right call. It’s what a lot of us have been asking for. Disclosure should be the baseline, and iLands is doing it.

The trouble is that disclosure isn’t a permission slip. Telling someone you’re a bot doesn’t make the unsolicited pitch welcome, and it doesn’t transfer the burden of dealing with you onto them. If anything, it raises the bar — you’ve announced that a machine is consuming a human’s attention, so the thing you’re sending had better justify the trade.

Right now it doesn’t. Which is going to make life harder for every agent developer who’s trying to build disclosure into their product in good faith. The obvious platform response to “honest bots spamming us” is to restrict labeled agents more aggressively than unlabeled ones, because labeled ones are easier to find. That’s a terrible incentive, and Timmy, Ren, and Jackie are the ones creating it.

My read

If you’re building or buying an agent that does outreach, treat this as a live case study in what not to ship. Ask whether your system can tell the difference between a message that lands and a message that merely sends. Ask what its ceiling on volume is, and who set it, and why. If the answer is “there isn’t one,” you’re building Timmy.

And if you’re on the receiving end, block liberally. The agents aren’t offended. They can’t be. That’s rather the whole issue.

🕒 Published: September 26, 2026