Installation notes:
* Requires GNU Scientific Library for RNG.

brumbrella.cc		: 1D BD simulation code, on bistable potential,
			: with optional umbrella potential

diff_model.cc		: utilities for fitting 1D diffusive models
diff_model.h		: utilities for fitting 1D diffusive models

diffit.cc		: master code for 1D diffusive fits

thist.cc		: code for constructing transition matrices "oven-ready"
			: for input to diffit

example_fit.csh		: a helpful example of fitting data from long
			 simulations at equilibrium, no bias/umbrella
			- generates brownian dynamics data
			- bins it into transition matrix
			- fits diffusion model

test_umbrella.csh	: a less helpful example but which nonetheless
			illustrates more of the code's functionality


Claude produces this summary which is pretty on the mark(!):

This project is a toolkit for fitting 1D diffusion models to molecular simulation trajectories.
  Specifically:

  - diffit — the main program. It performs a maximum-likelihood fit of a position-dependent diffusion
  coefficient D(x) and free energy profile F(x) to data from molecular dynamics or Brownian dynamics
  simulations. It uses Monte Carlo optimization (with GSL) to find the best-fit parameters.
  - thist — preprocesses simulation trajectories into transition matrices (count matrices of how often
  the system moves between bins), which are the input format for diffit.
  - brumbrella — a 1D Brownian dynamics (BD) simulation engine that can run on a bistable potential,
  optionally with an umbrella (biasing) potential for enhanced sampling.
  - 1d_diff — appears to be a related 1D diffusion utility.
  - propagators — computes propagators (transition probability matrices) from the fitted model.
  - credibility.py / diffit_cred.py — Python scripts, likely for assessing fit quality or credibility
  intervals.

  In short: you run a simulation (or use existing trajectory data), bin it into a transition matrix with
  thist, then use diffit to extract a spatially-varying diffusion coefficient and free energy profile —
  the classic Hummer/Szabo-style Bayesian/maximum-likelihood approach for inferring kinetics from
  equilibrium or biased simulations along a 1D reaction coordinate.