How Research Agentic Systems Are Actually Built: From ReAct to the AI Scientist

Important notice

Participants are to read the material in advance to engage more fully with the technical and methodological details during the session.If you have your own research, experience or saw something on the internet (blogpost, article etc.) that would add to the discussion, do contribute!

AI agents have evolved from simple reasoning loops into systems that can query external knowledge, call external tools, learn from failure, write and execute code, and now attempt increasingly end-to-end research pipelines. This session cuts through the hype to focus on the design ideas that made this possible, and what it means when AI systems begin participating in the research process itself.

The main paper is "Towards End-to-End Automation of AI Research" (Nature, 2026), supported by four lineage papers that show how the field arrived here: ReAct for reasoning grounded in action, Toolformer for autonomous tool and API use, Reflexion for learning from failure through verbal feedback, and SWE-agent for operating in real code execution environments. Together, these four papers capture the capabilities that The AI Scientist brings together into an end-to-end research pipeline.

The session will also critically engage with what the main paper actually achieves and where it falls short—including its authors’ concerns about research integrity, review-system overload, and the future of scientific training.

Suggested pre-reading:· Main paper: Towards End-to-End Automation of AI Research · Lineage papers: ReAct, Toolformer, Reflexion, SWE-agent· Optional (industry and forward-looking context):  OpenAI Deep Research System Card, Anthropic's How We Built Our Multi-Agent Research System, Google AI Co-Scientist, OpenAI PaperBench, Anthropic Automated Weak-to-Strong Researcher

More About the Host

Anshu Singh is an AI and Data Privacy Research Engineer at the Government Technology Agency (GovTech), Singapore. Before joining GovTech, her research focused on the intersection of computer vision and privacy at the NUS Centre for Research in Privacy Technologies. She holds a Master’s degree in AI from the National University of Singapore and enjoys building practical, user-centric solutions by putting research into practice.

Keshav Nath completed his Master of Computing from the National University of Singapore in 2026, after earning a Bachelor’s degree in Mechanical Engineering from Delhi Technological University. He has experience across both applied AI and research. He was a student researcher at the UCLA Smart Energy Lab and completed the fully funded MITACS Globalink Research Internship onsite in Toronto, resulting in three publications. In Singapore, he worked as a Junior AI Engineer at Staple AI, building production-grade agentic systems, and as a Graduate Teaching Assistant at NUS for Advanced Analytics with Big Data.

More About the Series

​Paper Club is Lorong AI’s community-driven initiative where members gather to discuss and analyze academic papers, research articles, or key developments in artificial intelligence.

​Get involved: Learn more about Lorong AI | Speaker Sign-up | WhatsApp Community | LinkedIn | X