Today's edition highlights the growing trend of automated harness discovery and self-evolving multi-agent systems, alongside a strong showing from open-source repository evolution tools.
The loop: MILO mutator agents rewrite complete agent harnesses and receive parent-specific feedback on their performance. An orchestrator uses this global search history to adapt the mutators' assignments and curriculum, improving the discovery of future harnesses.
A framework that co-evolves agent harnesses and the multi-agent strategy used to discover them by orchestrating mutator agents.
The loop: TOCOMAS proposes coupled changes to its own agent, collaboration, and memory policies during online execution. It retains candidates that satisfy structural constraints and improve evaluated reward, directly evolving its own architecture for future tasks.
A multi-agent system that enforces topological coherence across task dependencies and self-evolves its agent, collaboration, and memory policies.
The loop: The B-OPSD policy temporarily trains ahead to create a stronger future teacher. This teacher then generates reliable trajectories and provides dense supervision to the restarted original student, improving the model's own successor.
A self-distillation method that temporarily trains a model ahead to create a future teacher that provides dense supervision to the restarted student.
The loop: The X-Tree tokenizer mines flat action streams to build a hierarchy of reusable skills without LLM calls. This tree is then used as a self-teacher during on-policy self-distillation, improving the agent's ability to generalize across tasks.
A tokenizer that processes flat action streams into a reusable experience tree to guide efficient agent generalization.
The loop: ThumbGate Pre-Action Checks analyze ranked lessons and repeated failures from past executions. The system uses this data to self-improve its strict mode blocking rules, becoming better at preventing secret leaks in future actions.
A pre-action check system that self-improves its strict mode blocking rules by learning from past execution failures.
The loop: Proteus plugs into any agent harness to measure its performance and propose evolutionary changes. These changes are applied to the harness, improving the execution environment for subsequent agent runs.
A plug-and-play framework designed for the self-evolution and continuous measurement of any agent harness.
The loop: The Dream Machine engine schedules nightly repository evolution tasks and evaluates the results. It uses an evidence-gated promotion system to merge successful changes, continuously improving its own configuration and codebase.
A configuration-driven engine that schedules nightly tasks for evidence-gated repository evolution and self-improvement.
The loop: The AutoSynthData generator creates synthetic training datasets tailored for enterprise environments. This data is then fed back into the training pipeline to improve the performance and reliability of the enterprise agents.
A tool that generates synthetic training data to enhance the performance and reliability of enterprise AI agents.
