Given that he doesn't seem like a soup-kitchen kind of guy (from reading his Medium), why not take a new approach where you try one thing, understand one person in depth, see what you can do, and work from there? That's how you learn and get in-touch with anything.
There's more than one way to skin a cat. People do methods A, B, and C all the time. This guy is trying method D, and doesn't seem to be gunning for repeatable or sustainable or "solve the homeless problem in one fell swoop". He comes across as arrogant and out-of-touch on his blog but he's basically got the dude's back. IMO, at the end of the day all of that adds up to a good thing.
Also, I've worked with some NGOs and I've found them to be generally ineffective at solving problems but great at making volunteers feel good about themselves so the last place I'd want to go if I wanted to understand homelessness would be aid workers, and the first place would be some reading, my own instincts, and working with one individual homeless person. I personally like the "ballsy n=1 experiment" approach he's taking - even though his language makes him come across as kind of a dick.
Also, he is making a real sacrifice - one in terms of time/effort teaching him to code, and two in terms of reputation/job risk. If Leo fails, or if the media twists it the wrong way, McConlogue looks like a jackass and he seems smart enough to understand that.