A workshop organized under Argonne National Laboratory researcher Lois Curfman McInnes's 2024 DOE Office of Science Distinguished Scientist Fellowship has published a report identifying four strategic themes for scientific computing as AI shifts from tool to active research participant.

The report, "Toward Next-Generation Ecosystems for Scientific Computing: Harnessing Community, Software, and AI for Cross-Disciplinary Team Science" (L. C. McInnes et al., available at https://doi.org/10.48550/arXiv.2608.26519), grew out of the second workshop in a three-year series, held April 14–16, 2026, in Chicago. Researchers and practitioners from universities, national laboratories, industry and other organizations attended. Where the 2025 workshop identified challenges, this year's sessions moved toward strategic priorities and community actions.

The four themes are interdependent: software ecosystems for AI-enabled scientific discovery; trust, validation and traceability; human-AI teaming and paradigm shifts; and workforce, pedagogy and governance.

A broad definition of ecosystem. The report uses "ecosystem" to cover technical, social and organizational structures — hardware, software, data, workflows, people, institutions, incentives and community practices. The goal is an environment that adapts as scientific needs and technologies change. As AI systems increasingly generate and analyze code, synthesize information, assist with hypothesis generation and carry out complex reasoning within scientific workflows, questions of validation, provenance, accountability and the role of human expertise become more pressing, the report argues.

Shared research assets. Effective software ecosystems require more than access to computing facilities, the report states. Shared research assets include reusable software, datasets, metadata, workflows, benchmarks, validation examples and documentation — forms that capture knowledge, context, assumptions and computational approaches for reuse across projects and communities. These assets help researchers understand, extend and validate complex workflows as AI, accelerator-based computing and emerging technologies become increasingly interconnected. The report also stresses usability: helping researchers identify appropriate methods and their limitations.

Trust as a design requirement. The report calls for verification, validation, uncertainty quantification, provenance, traceability, transparency and auditability to become core design requirements rather than afterthoughts. Traditional verification and validation approaches may fall short as AI-enabled workflows move from exploratory use to scientific analysis and decision-making; they must extend to hybrid workflows combining mechanistic models, numerical methods, AI systems, surrogate models and human judgment.

Scientific software poses a particular problem: it can run without a technical error and still produce scientifically misleading results. Conventional tests may not catch this, especially when AI rapidly generates or modifies code. Workshop participants stressed the need for new mathematically and statistically grounded approaches to evaluating AI-enabled scientific workflows.

From taskwork to teamwork. The report distinguishes taskwork — activities needed to reach a goal — from teamwork, which concerns how people and AI systems coordinate toward it. Adding more agents to a task does not automatically improve a workflow. Human researchers remain essential for framing questions, combining domain and computational knowledge, evaluating evidence, managing uncertainty and interpreting unexpected results. The right balance between automation and human involvement depends on the task; higher-risk activities demand greater transparency, validation and human judgment. The report recommends giving early-career researchers practical human-AI teamwork experience through fellowships, internships, apprenticeships and cross-sector projects.

Workforce for continual change. Training must extend beyond programming to critical thinking, evaluation of evidence, reasoning under uncertainty, verification, reproducibility, communication and responsible use of AI-enabled tools. The report's workforce roadmap offers multiple pathways for students, practitioners and institutional leaders, aiming to build competencies that outlast any generation of tools. It also flags a need to understand how AI may reshape scientific roles, institutions and career paths.

Community action. The central message: challenges intensified by AI cannot be solved by individual projects or organizations. Coordination among researchers, universities, national laboratories, the private sector, funders and professional societies is required. The report identifies eight action areas. Near-term efforts should strengthen shared research assets, trust infrastructure, human-AI teaming practices and incentives for scientific software stewardship. Pilot projects should explore coordinating multiple AI agents and institutions, and investigate helping users make better-informed choices throughout research workflows. Longitudinal studies should examine how AI affects the workplace, professional roles and team behavior, and evaluate whether AI-enabled ecosystems actually improve scientific practice.

The report frames the stakes plainly: AI can accelerate parts of scientific work, but speed alone is not progress — the question is whether AI-enabled ecosystems support valid, interpretable and creative discovery. "Done well, these ecosystems can accelerate discovery and expand scientific inquiry while preserving the rigor, transparency, accountability and collaboration on which progress and public benefit depend," it concludes.

Attendees of SC26 can continue the discussion Tuesday, Nov. 17, at the birds-of-a-feather session "Ecosystems for Scientific Computing in the Age of AI." The third workshop in the series will determine whether the community's near-term actions — on shared assets, trust infrastructure and multi-agent pilots — translate into measurable improvements in scientific practice.