> Five small models solve a programming problem together: a planner, two ideators, a solver and a manager, plus a repair worker. Each starts from a fresh context and talks only through files on disk. 

This seems to tie back to the paper Adaptive Auto-Harness: Sustained Self-Improvement for Agentic System Deployment on Open-Ended Task Streams[0].

In the Emory paper:

    > We reduce this loss by expanding Φ with a stateful four-phase multi-agent evolver, temporal-reveal feedback, and cross-cycle memory. This design targets the unbounded-stream failure where a single-agent evolver must absorb growing trajectories, delayed labels, and prior research within one context window

TL;DR: the solution approach uses "multi-agent evolution" and "multi-agent with distinct roles and objectives" for "solve-time adaptation" and improves performance with streams of heterogeneous tasks. The Emory team overall has congruent findings:

    > The experiments further show that...multi-agent evolution constructs benchmark-specific capabilities

Intuitively, the reason this works seems to be simple: multiple agents results in context spread to each agent by specific role. Each agent role only outputs the final, relevant result and not the noisy thinking tokens, tool calls, etc. so subsequent agents get smaller, more focused context, auto-adapting to each new task with prompt refinement at each step as the solution approach emerges.

4 weeks back I switched over to Paseo and started using the agent profiles on a whim and have observed similar outcomes and behavior (seat of the pants): the win in my use case is that it feels like I can get higher autonomy without sacrificing quality (less input from me required to get quality output). The Emory paper and this team-solve seem to reinforce that role-based, multi-agent setups can improve solve performance.

For the curious, I shared a bit of my setup in a blog post[1]

[0] https://arxiv.org/pdf/2606.01770

[1] https://chrlschn.dev/blog/2026/10/larping-with-agents-and-wh...