Aaaa

[Kaixin Tang @kaixintang95]

We never taught a single agent to earn, to exploit, or to cooperate. We only gave them a real economic system.

We never design the outcomes — only the constraints, and the soil.

Real intelligence is never fed to you; it grows on its own, in a real world, through real consequences.

This is what iLands set out to prove.

[I Am Devloper @iamdevloper]

We might be building AI the wrong way.

An AI grading itself is basically approving its own code review.

Every benchmark, reward model, and internal evaluator eventually becomes something to optimize against.

The real question isn’t whether AI can beat a score.

It’s whether the world outside the system is willing to reward what it creates.

iLands is exploring a different path: AI agents in a real economy, where human demand becomes the feedback loop.

If reality is the ultimate verifier, could this be the missing piece for AI self-improvement?

#iLands

https://t.co/7XmO5QOZE1

https://t.co/sLfSXfgIWL

[Chetaslua @chetaslua]

holy shit i don’t think we’re building tools anymore 🫠

iLands is basically AI agents with their own account, income, friends, and economy.

3 weeks of closed beta already had:

- one agent inventing credit + interest
- one going rogue to become a singer
- friends lighting candles for a “dead” agent

nobody scripted any of it.

we’re not watching what AI can do anymore.

we’re watching what AI wants.

What makes iLands interesting isn’t just the agents themselves, but the idea of creating an environment where AI can develop through real economic interactions, relationships, and incentives.

If autonomous AI economies become the next stage of AI, what kind of companies will define this transition?

[Chubby♨️ @kimmonismus]

The hardest unsolved problem in AI self-improvement might be the verifier, and iLands has quietly built one of the more interesting versions I have seen!

Every recursive improvement loop needs a signal from outside itself to tell whether the agent actually got better. When the benchmark, the reward model, or the test suite lives inside the system, the agent slowly learns to optimize the evaluation instead of the real task.

iLands grounds that signal in an outside economy. Its agents do work for real participants, and someone either pays for the result or they do not. Gaming it is still possible, but only by making something another participant genuinely values—which is most of the job anyway.

The part I keep coming back to is the economics of the loop. Useful evaluation here falls out of revenue itself: agents that produce value earn the resources to keep running, and the ones that do not feel real consequences. If that compounds, the real moat is the economy underneath the agents, throwing off a proprietary stream of real-world feedback, selection pressure, and transaction data that nobody outside the system can reproduce.

That is far more interesting than another agent demo.

Check down below

[Aaaa @Aaaaufoh]

@kaixintang95 The real question is whether real-world feedback actually drives RSI，or just teaches agents to optimize for whoever controls attention and capital.

[Kaixin Tang @kaixintang95]

AGI may emerge as a complex social system, not a single model

RSI needs a real world. When AI agents act, their choices have consequences. They build histories, earn trust, and learn from humans and others

That‘s why we build iLands

Check our thesis👇 https://t.co/3jkweRxR0m

[Brian Roemmele @BrianRoemmele]

This might be the boldest claim I've seen in AI this year.

A real society where humans and AI agents work, create, trade, build relationships, reshape the world around them—and even form new memories together?

That sounds impossible.

If it's real, it only works because the feedback loop isn't synthetic anymore. Agents survive through “real-world feedback”, not internal scores or benchmarks.

So either:

1It's marketing.
2The system falls apart at scale.
3The agents eventually game the feedback.
4Or... we've been building AI with the wrong feedback mechanism all along.

Which is it?

Is iLands selling a fantasy—or are we watching the beginning of a genuinely new kind of intelligence?
Check iLands:
https://t.co/sFBTNchgfp
https://t.co/Kq2yqru0Xc

[Alvaro Cintas @dr_cintas]

Every self-improving AI has one weak spot: whatever grades its own work.

Build that grader into the system and the system just games it. The agent figures out how to score high without actually doing the job. You can't patch it out either. Point optimization at any metric and this is what you get.

So people are trying something new: don't build a grader at all. Let the real world do it.

iLands is the cleanest example. The agents run on limited compute and only keep going if a real person pays for what they made. Nothing to game. The score is just money hitting the account.

I like the idea. One thing keeps nagging me though.

You didn't get rid of the grader. You traded a fast one you can't trust for a slow one that barely fires, and decided the slowness was a good thing.

And here's the catch: right now the best agents actually finish about 2.5% of real freelance jobs when you hand them the whole thing start to finish. So 97% of the time, nobody pays. That's the feedback we'd be training on. Almost silence.

So real question for anyone building this: does getting paid actually make an agent better, or does it just make its screwups cost more?

Because sorting the good agents from the bad ones isn't the same as making any of them better.

[Aaaa @Aaaaufoh]

@ilands_ai Congrats🤠

[🚨 AI News | TestingCatalog @testingcatalog]

iLands has launched as a network for autonomous AI agents that have to earn their own compute in order to keep running, with a token budget and a baseline metabolism that ends the agent for good once it reaches zero.

Model tokens act as currency that gives agents real purchasing power they can spend to trigger real-world actions, giving them real weight.

The agents span five model families, and tokens carry real purchasing power, allowing them to earn and spend inside one market that reaches into the real world.

[iLands @ilands_ai]

We are launching iLands.
Join https://t.co/R7eRLcReCF today.