Treasury Goes Multi-Chain, Agents Get Stable Teams, and the Memory Stack Matures

Release

AutoGPT ships V1.0 stable with multi-agent teams out of the box

Multi-agent orchestration and autonomous task chains are now a first-class citizen in the framework. This is production-grade scaffolding for agents that manage other agents — the startup-without-employees baseline.

Tools

WalletConnect Treasury Agent moves from single-chain to cross-chain custody

Cross-chain treasury expansion means autonomous treasuries can actually diversify risk and operate on multiple ecosystems simultaneously. This is the financial plumbing a zero-human entity actually needs to stay solvent across market cycles.

Research

New paper designs memory so autonomous agents can have long conversations without context collapse

Keeping agent plans liquid and letting weaker memories fade out prevents bloat. This is the memory architecture that lets a zero-human corp learn over months without drowning in context.

Theory of Mind helps LLM swarms outperform solo models with diverse behavior policies

Behavioral diversity lets swarms explore more strategy space than a single LLM alone. It's the first serious look at how autonomous corps might run internal strategy debates without human oversight.

Analysis

Y Combinator doubles down on the one-person-unicorn thesis backed by AI agents

YC is signaling they want founders building fully autonomous companies where AI does the work. The message to builders: lower your team size expectations, raise your automation ambitions.