SakanaAI/mass

Multi-Agent Self-Supervision (MASS)

Self-supervision through better coordination among agents that share one language model.

Sakana Fugu — Multi-agent System as A Model

One model to command them all

Recursive Self-Improvement through Multi-Agent Self-Supervision

Recursive self-improvement (RSI) of a model on non-verifiable tasks, such as open-ended research, faces a supervision bottleneck when its outputs exceed what even human experts can reliably assess, leaving the model itself (optimizee) as the best available optimizer and evaluator. However, a single model instance struggles to critique and improve its own complex reasoning under this homogeneous loop. To address this, we propose Multi-Agent Self-Supervision (MASS), an RSI method that alternates between evolutionary workflow optimization and supervised fine-tuning on self-generated trajectories. Guided by early findings that multi-agent topologies excel at complex reasoning, MASS prompts a single base model to iteratively propose, execute, and self-evaluate multi-agent workflows. Through an evolutionary search constrained by structural guardrails, the model optimizes these computational-graph-like orchestrations, discovering the most effective distinct roles and information routing for a given task. Over two MASS cycles with Qwen3.6-27B, the model achieves 1.2-1.6x higher performance per output tokens on four open-ended public benchmarks. Because the improved model subsequently acts as a better optimizer and evaluator, this alternating framework enables a continuous, recursive bootstrapping of the model's capabilities. Moreover, multi-agent traces are also more training-efficient: a student trained on them outperforms a single-agent student trained on 1.4x more training tokens. These findings suggest that jointly learning orchestration and bounded subagent execution from multi-agent trajectories can provide an effective signal for RSI.

LukeW | Agent Management Interface Patterns

As an increasing number of AI applications evolve to agents doing work for people, agent management becomes a critical part of these product's design. How can p...

AI agent runs amok in Fedora and elsewhere

Agentic AI systems can be used to do a variety of things autonomously on behalf of a human user [...]

Agents as scaffolding for recurring tasks.

One of my gifts/curses is an endless fixation with how processes can be optimized. For a brief moment early in my career, that was focused on improving how humans collaborate, but that quickly switched to figuring out how we can minimize human involvement, and eliminate human-to-human handoffs as much as possible. Lately, every time I perform a recurring task–or see someone else perform one–I think about how we might eliminate the human’s involvement entirely by introducing agents. This both has worked well, but also worked poorly, and I wanted to highlight the pattern I’ve found useful.

Agents are actors

Multi-agent is just actor model

Dii familiares

Musings on agents and AI and a little signposting

Structured reasoning about actor systems | Proceedings of the 2013 workshop on Programming based on actors, agents, and decentralized control

Successfully attaining consensus in the absence of a centralized coordinator is a fundamental problem in distributed multi-agent systems. We analyze progress in the Synod consensus protocol—which does not assume a unique leader—under the ...

Structured reasoning about actor systems | Proceedings of the 2013 workshop on Programming based on actors, agents, and decentralized control

Successfully attaining consensus in the absence of a centralized coordinator is a fundamental problem in distributed multi-agent systems. We analyze progress in the Synod consensus protocol—which does not assume a unique leader—under the ...

MemGPT: Towards LLMs as Operating Systems

Letta is the platform for building stateful agents: AI with advanced memory that can learn and self-improve over time.

Managing Agent Skills with Your Package Manager | pavel.pink

AI coding agents use skills — markdown files that teach them domain-specific tasks. We publish ours as conda packages and manage them with pixi, getting versioning, lockfiles, and supply chain security for free.

Group Overview ‹ NANDA – MIT Media Lab

NANDA: The Internet of AI Agents

The Agent Harness

A specification for agent behavior that LLM frameworks leave undefined: error handling, context management, tool execution, and state transitions.