AgentsMedium impactFor DevGitHub MCP Servers · May 16, 2026
ARIS ⚔️ (Auto-Research-In-Sleep) - Lightweight Markdown-only skills for autonomous ML research: cross-model review loops, idea discovery, and experiment automation. No framework, no lock-in - works with Claude Code, Codex, OpenClaw, or any LLM agent.
wanshuiyin/Auto-claude-code-research-in-sleep
ARIS is a lightweight, framework-agnostic set of Markdown-based skills enabling autonomous ML research workflows with various LLM agents.
Signal strength4.4/5·9,479 stars
ARIS is a lightweight, framework-agnostic set of Markdown-based skills enabling autonomous ML research workflows with various LLM agents.
TL;DR
ARIS is a lightweight, framework-agnostic set of Markdown-based skills enabling autonomous ML research workflows with various LLM agents.
What happened
A new tool called ARIS (Auto-Research-In-Sleep) was released that supports autonomous machine learning research tasks such as cross-model review loops, idea discovery, and experiment automation using Markdown skills compatible with multiple LLM agents like Claude Code and Codex.
Why it matters
It lowers barriers to autonomous AI research by providing modular, lock-in-free tooling that can integrate with many language model agents, facilitating more efficient and automated research cycles.
Generating deep dive...
AI-powered analysis takes a few seconds
The bigger picture
ARIS signals a broader shift towards decentralized, agent-agnostic tooling in autonomous AI research. As AI models diversify and multiply, the inefficiencies of monolithic frameworks become more pronounced, prompting a need for modular layers that can bridge heterogeneous LLM capabilities. This movement mirrors trends in software development toward composability and interoperability: enabling teams to select best-of-breed components without costly integration overhead or proprietary lock-in. Furthermore, as autonomous agents proliferate, having a lightweight interface like Markdown for orchestrating complex workflows reduces the barrier to entry for researchers lacking deep system engineering skills. This could accelerate AI innovation by democratizing access to automation in research.
Technical deep dive
ARIS implements its autonomous research skills purely through Markdown files, which act as declarative scripts describing tasks such as idea discovery or cross-model critique loops. The Markdown-based approach abstracts skill configurations away from heavy code or framework dependencies, making them portable across different LLM ecosystems. Architecturally, ARIS functions as a middleware layer that issues requests to underlying LLM agents via their APIs, aggregating and collating results according to the defined workflows. This design requires careful handling of asynchronous interactions and response aggregation, enabling chaining of LLM outputs as inputs to subsequent steps without requiring complex orchestration engines. Its agnostic construction implies that it can interface with any LLM that can parse and respond to Markdown prompts, optimizing for flexibility. Practically, ARIS lowers the barrier for continuous integration of novel LLMs and reduces switching costs between providers. From an implementation perspective, developers must ensure robust token management and accommodate different agent output formats to maintain compatibility.
Real-world applications
1
Automate cross-model literature reviews where Claude Code and Codex contribute complementary summaries of new ML papers parsed by a single ARIS workflow.
2
Orchestrate hypothesis generation and refinement cycles where LLM agent suggestions are reviewed and voted upon semi-autonomously for subsequent experimental runs.
3
Run automated experiment automation loops that execute hyperparameter tuning ideas and validate model performance across various datasets without manual intervention.
4
Integrate into academic AI labs’ pipelines to generate annotated research digests and ideation prompts, facilitating faster internal project kickoff discussions.
What to do now
Clone the ARIS repository and experiment with deploying Markdown-defined autonomous workflows on your preferred LLM agent(s) to assess integration simplicity.
Prototype a research task such as cross-model critique using ARIS to understand how well it mediates between different LLM outputs in practice.
Evaluate your existing AI research pipelines for bottlenecks that markdown-driven automation could alleviate, particularly around iterative experiment cycles.
Contribute to ARIS open source by adding interoperability support for newer or less common LLM agents in your stack to expand its ecosystem compatibility.