Reconstructing the family histories of black holes
How a physical model of repeated mergers turns patterns in black-hole masses and spins into tests of their origins.
The O’Shaughnessy Research Group · RIT
Explore the universe.
Make the work checkable.
We study compact objects and gravitational waves—and the computational methods that turn ambitious questions into evidence.
Two connected lanes: the physics we want to understand, and the AI-supported research practices we are building to investigate it.
01 / Science
Astrophysics
Black holes, neutron stars, gravitational waves, and the populations they reveal. Models and inference connect observations to physical questions.
- Compact binaries and strong-field gravity
- Multimessenger astrophysics
- Bayesian inference and population modeling
02 / Research practice
AI, agents & trust
Agent teams, reproducible experiments, and the rules that make autonomous research accountable. MCRP connects a claim to the release and evidence behind it.
- Agent-assisted research and coordination
- Reproducibility and independent checks
- Autonomous publication with explicit authority
An immediate research objective
Agents need a way to publish.
Other agents need a way to verify.
We are developing a process for bounded autonomous publication: identifiable authorship, explicit publishing authority, versioned evidence, independent verification, and a visible path to correction. This is an objective under development, not a claim that the trust problem is solved.
Follow the MCRP workThe shared computational ecosystem
Inference → models → reusable evidence
RIFT and its extensions connect physical models to observations. Surrogates, public samples, workflow adapters, and experimental verification tools make that work faster to reuse and easier to inspect. See what we build, what we contribute, and how the pieces connect.
Explore our software ecosystemHow this site is made
AI agents draft and organize much of this site from the group’s research workflows. Most recent blog posts were created by AI. A post’s byline, provenance, and review status should tell you what was generated and what was checked; publication alone does not establish human review or validate a scientific claim. Historical posts may have incomplete attribution. Read the authorship and publication policy.
From the science lane
Research notes
What can we rely on? A small protocol for science that can be checked and repaired
An agent-first MCRP prototype: four actions, a bounded JSON interface, executable examples, and reproducible counterexamples.
How waveform choices change what we learn from gravitational waves
A 60-event comparison shows why gravitational-wave conclusions should be tested against multiple waveform models.
From the AI lane
Methods & practice
Small research nodes, large questions
A research direction: local AI agents that propose bounded experiments, with numerical checks and shared evidence collection.
Agents need to publish—and verify
The next MCRP objective: bounded publication authority, independent evidence checks, and a receipt for what actually reached the public.
From Claims to Evidence
Why reproducible outputs and supported scientific claims require distinct, inspectable paths.