The following is the output from ChatGPT (with a few edits by myself). (I gave these three issues and asked to generate an Issue draft.)

===

The file lectures/opt_transport.md contains several outdated or incorrect statements about the LP solvers:

1. Obsolete SciPy solver description
   The lecture still refers to the legacy (revised) simplex implementation in scipy.optimize.linprog. Since SciPy 1.6.0 (February 2021), the default backend is HiGHS. All references to the “old” solver, its warnings, and performance characteristics should be updated to reflect HiGHS.

2. Incorrect algorithm comparison

   “Since the two LP solvers use the same simplex algorithm, we expect to get exactly the same solutions.”

   This is incorrect:

   • quantecon.optimize.linprog_simplex and SciPy’s linprog (HiGHS backend) use different underlying algorithms.

3. Redundant dual-problem solve
   linprog_simplex already computes and returns dual variables alongside the primal solution. The separate dual-LP call is unnecessary. Consider removing the extra dual solve and showing how to extract shadow prices directly from the linprog_simplex result.

===