My one caveat with that approach is it requires a very specific implementation of their library in MIT Scheme. There probably is a racket implementation by now, but it still looks like it's most small personal github projects.
The automatic differentiation landscape has changed considerably since that book was written. I would love to eventually be able to put together something like a guide to Lagrangian mechanics with JAX.
As a counterpoint, I am someone who did self-learn physics from Feynman's lectures as a teenager and I would highly recommend it to others to at least give it a try.
You have to appreciate just how much of a factor pure interest and enthusiasm for a subject is in learning it - and Feynman's lectures are just amazing for sparking that joy in physics. I discovered Feynman's lectures about two months into my first year of high school and it became my primary textbook. I just found the other texts too goddamned boring in comparison.
However I would not recommend the Feynman lectures (and, based on the preface, I think Feynman would agree). I've read about a 1/3 of the series and while I have to say they are some of the greatest technical writing in history, they are not the greatest pedagogical material. I do believe everyone should read at least some of the Feynman lectures, just not as a means of learning physics. Much like a good 3blue1brown video, you feel like you're learning a lot... but 6 months later there's just a big empty hole were knowledge used to be.
Based on recommendations I've found I'm currently working through Taylor's Classic Mechanics. My ultimate aim is to get a better grip on Lagrangian mechanics and make use of the plethora of great automatic differentiation tools out there to create some trick simulations of physical processes. So far the book has been fantastic, but I'm not far enough through to be able to give it an unequivocal endorsement.
Of course self learning does miss out on the experimentation portion, but I suspect a software engineer could afford the basic tools to setup a variety of high-school/early undergrad experiments. If you're hankering for something requiring institutional funding level of equipment you should be able to audit a course at your local university.