ORCD Newsletter: November 2025

IAP 2026 Classes

  • Build a Mini-SuperComputer
    • January 13, 14, and 15, 1-4PM (NE45-200)
    • In this hands-on IAP course, you’ll design and build a working supercomputing cluster using Raspberry Pi components.
    • See the MIT calendar listing for more information and registration.
  • Practical Computational Thinking
    • ORCD together with MIT Libraries, and Schwartzman College of Computing will be offering a hands-on sequence of classes “Practical Software Carpentries,” “Practical Programming with Data,” and “Practical High-Performance Computing” that cover skills from beginners to advanced skills useful for using computing in research.
    • Scheduling information will be posted on the ORCD website in the coming weeks.

Cybersecurity and Artificial Intelligence

A new report from Anthropic detailed an Artificial Intelligence (AI) attack using their Claude platform that allowed AI agents to attack a large number of targets. The report says “...the human operator tasked instances of Claude to operate in groups as autonomous penetration testing orchestrators and agents, with the threat actor able to leverage AI to execute 80-90% of tactical operations independently at physically impossible request rates.” Anthropic was able to thwart the attack by disabling Claude accounts. This brings a new dimension to cyber security at MIT and beyond to think about.

Seed Fund Updates

We’ve selected eight more Seed Fund projects from the 35 applications we received for the spring 2026 semester. This is round two of the ORCD seed fund program. We will continue opening up new rounds of funding in 2026 and 2027, announcing new application launches on our website and in our newsletter. The new projects span all parts of MIT and range from exploring urban insect population dynamics to improving neural network back propagation efficiency.

Meantime, our inaugural round of ORCD Seed Fund projects is wrapping up. During the summer and fall of 2025, eight projects were given access to enhanced GPU resources and research assistant funding to pursue their research while providing the ORCD team with feedback and testing on our systems, including some experimental tooling. Projects came from Chemical Engineering, CSAIL, EECS, the Kavli Institute for Astrophysics and Space Research, Mechanical Engineering, the Media Lab, and the Schwartzman College of Computing.

Our first group of projects included:

  • a research assistant platform designed to deliver rapid and precise responses to complex queries using LLMs in the field of gravitational-wave physics
  • an open-source benchmark to study risks during the pretraining phase of LLMs
  • a universal, interactive 3D segmentation system for biomedical images that can dramatically reduce the time clinicians and biomedical researchers spend manually segmenting 3D volumes
  • extending a newly developed fractal-based generative modeling framework to scientific domains where data exhibit self-similar, hierarchical structures
  • accelerating enzyme simulation with a biochemical hierarchy of neural network potentials
  • a large-scale foundation model capable of reasoning across multiple medical modalities
  • a novel video-based robot foundation model that integrates advanced generative modeling with robotic control for manipulation, navigation, and general scene simulation
  • a foundation model for high-resolution satellite imagery

Wei Dai recently presented his Seed Fund work with Paul Liang and others at NeurIPS. Other projects are currently writing up results for publication and sharing.

December and January Office Hours

ORCD office hours are cancelled on Tuesday, December 23; Thursday, December 25; Tuesday, December 30; and Thursday, January 1.

What We’re Reading