@sydneyrunkle: we just shipped support for rubrics in deepagents give your agent a clear definition of what "done" looks like, and for…
Summary
Sydney Runkle announced support for rubrics in deepagents, allowing agents to define a clear definition of done and loop until the goal is complete.
View Cached Full Text
Cached at: 06/09/26, 10:45 AM
we just shipped support for rubrics in deepagents ✅
give your agent a clear definition of what “done” looks like, and force it to run in a loop until said goal is complete
this is similar to /goal in claude code, but works for any agent (not just a coding agent) https://t.co/FMjBubuf4h
Similar Articles
@sydneyrunkle: https://x.com/sydneyrunkle/status/2071629451712983319
Deep Agents introduces dynamic subagents that use programmatic orchestration via code scripts instead of tool calls, enabling reliable scaling and complex workflows. The feature integrates with a QuickJS code interpreter for lightweight execution.
Co-ReAct: Rubrics as Step-Level Collaborators for ReAct Agents
Co-ReAct introduces a rubric-guided action-selection framework that uses rubrics as step-level guidance during inference for ReAct agents, improving trajectory quality and outperforming baselines on DeepResearchBench and SQA-CS-V2.
DuMate-DeepResearch: An Auditable Multi-Agent System with Recursive Search and Rubric-Grounded Reasoning
This technical report introduces DuMate-DeepResearch, a multi-agent framework for deep research tasks that decouples the agent core from a tool ecosystem, and incorporates graph-based dynamic planning, recursive two-level execution, and rubric-based test-time optimization. The system achieves state-of-the-art results on two deep research benchmarks, demonstrating the value of auditable agent infrastructure.
@sydneyrunkle: people want to build agents, and they want it to be easy even more important, it should be easy for your agents to impr…
People want to build agents easily and have them improve over time; automating the 'hill climbing' loop is hard but has high ROI.
RUBAS: Rubric-Based Reinforcement Learning for Agent Safety
RUBAS is a rubric-based reinforcement learning framework for agent safety that decomposes LLM agent behavior into four dimensions—tool-use safety, argument safety, response safety, and helpfulness—providing fine-grained rewards over complete trajectories. Experiments show RUBAS improves safety over standard alignment baselines while reducing tool-grounded hallucinations and maintaining competitive utility.