The Dresher Center for the Humanities is continuing our writing group into the Spring 2026 semester on Fridays from 10am to 11:30am online via WebEx.

This writing groups open to faculty, graduate students, and staff working on humanities-related projects.

https://umbc.webex.com/umbc/j.php?MTID=m3e509a76259be1f8e3262d623e29b53d

The Mentor newsletter for Winter 2026 is now available!

 The Mentor is the MD State Retirement & Pension Agency's newsletter for Active members. The newsletter provides news, information, and legislative updates.

 The Winter 2026 edition of The Mentor features: 

• The importance of designating your beneficiary (use mySRPS)
• How to harness the power of secure messaging with mySRPS
• MSRA's new Chief Investment Officer (CIO)
• MSRA's 9.83% investment return for FY 2025
• How to use your military service to "boost" your future pension (you must apply)

The Maryland State Retirement and Pension System administers death, disability and retirement benefits for more than 427,000 members and retirees. 

Tanyka M. Barber (she/her)

Fabio Anza, an assistant professor in UMBC’s Department of Physics, Quantum Science Institute and Cybersecurity Institute, has won an Amazon Research Award to develop Physics Co-Pilot—a scientific AI assistant that combines large language model (LLM) orchestration with reliable computational algorithms to address critical bottlenecks in modern physics research.

The problem isn’t just coding. Theoretical physics, like many other areas of scientific research, is currently facing two major challenges: knowledge fragmentation and hyperspecialized computational tools. The open-access, online scientific paper repository ArXiv alone publishes over 25,000 new papers monthly across computer science, math, and physics. Meanwhile, using sophisticated computational methods requires mastering ultra-niche software ecosystems—work that diverts researchers from scientific discovery.

“Researchers spend increasingly large amounts of time wrestling with software issues rather than exploring scientific questions,” Anza explains. “Valuable human capital gets diverted from discovery to dealing with technical problems that could be avoided with better infrastructure.”

A hybrid approach to scientific computing

Physics Co-Pilot takes a fundamentally different approach from pure AI code generation systems. Rather than having language models write code—which, at scale, produces unreliable results—the system uses LLMs only for natural language understanding, while delegating all computational work to pre-written, validated algorithms.

Researchers spend increasingly large amounts of time wrestling with software issues rather than exploring scientific questions. Valuable human capital gets diverted from discovery to dealing with technical problems that could be avoided with better infrastructure. — Fabio Anza, assistant physics professor

“The LLM acts as an orchestrator, not a programmer,” Anza says. “It translates natural language into structured commands that call expertly implemented computational routines. This maintains computational rigor while making advanced techniques accessible through conversation.”

The tool features specialized software routines for literature analysis and physics simulations. For the initial release, the system will focus on modeling a a narrow set of physical systems, serving as proof-of-concept for future community-driven expansions.

From concept to open-source release

Anza has recruited computer science senior Samuel Truong to help develop the system over the next year, with plans to expand the team as the project progresses. 

The project follows a short timeline: an initial prototype with literature analysis capabilities will be released in the first six months, enabling researchers to conduct natural language-driven searches and synthesis of ArXiv papers. The complete system, including simulation and visualization capabilities for dynamical systems, will be released (under the MIT open-source license) by the end of 2026.

“Physics Co-Pilot represents a new paradigm in scientific assistants,” Anza notes. “By combining conversational accessibility with computational reliability, we aim to give researchers access to advanced computational techniques without the need to know any programming languages.”

The system aims to enable physicists to discover and synthesize knowledge across disciplines, eliminate programming barriers to sophisticated analysis methods, and accelerate research cycles through rapid hypothesis testing—all without researchers writing a single line of code themselves.