I've dug up a lot of papers on effectiveness of teaching strategies in physics in particular (since I'm a physics student), and it seems in physics that your time spent working practice problems and making predictions about experiments is more important than actual lectures.

An example:

http://www.sciencemag.org/content/332/6031/862.abstract

If anything, I think the Khan Academy is best suited for a mixed approach where students watch lecture videos, then attend a problem session or help session to work examples, get help on exercises, and so on. The Khan Academy's exercise framework is an important step in this direction, and I hope they continue to improve it.

Teletypes did have a one key, probably because they were all-uppercase, so there wasn't a lowercase ell available.

http://scratch.mit.edu/ http://info.scratch.mit.edu/sites/infoscratch.media.mit.edu/...

Then they work on doing something interesting while I answer questions and ask "What happens if...".

It's a lot better than some of in-person lectures I've taken, but maybe that's just me. I've been encouraging my peers to use online videos as a source to learn stuff they have a hard time grasping in class for quite a while. They don't listen, but that's another matter...

Anyways, I'm glad Khan Academy is coming out with this. I was freaking out when I couldn't solve some project Eular questions, and pretty much lost all of confidence in my programming abilities for a while. Online tutorials helped me a lot, and I think I've become a better programmer because of that.

One small nitpick: I haven't seen the videos yet, but it looks like they'll be in python. I'd rather it be C or C++. From what little I know of the language, it's pretty easy to pick up, and doesn't put pressure on the programmer to take care of details like memory management etc. I know for a fact that many algorithms that I've implementer in C++ that take over 100 LOC, I've found solutions online that have the same algorithm in 20-30 LOC. Used to irritate me a lot. :P

As experts, syntax doesn't annoy us. We see past the curly brackets (or the parentheses), because we can quickly see the internal AST that derive from them. Beginners can't.

Now, it would be wonderful to teach them a systems language that forces them to manage their own memory. I think a good candidate would be to keep most of the semantics of C, but simplifying its syntax (and find a solution to simplify includes).

There is a crucial point that should not be ignored. The first two languages someone learns should be easy. Monstrosities like C++ give the poor beginner the impression that learning any language is hard to learn, and that would discourage such learning in the future.

I disagree 100%, if you're talking about a college curriculum. The intro class at my university [1] used C, and I'm incredibly grateful. If I hadn't been forced to grok pointers and malloc, I'm not sure I ever would have (I haven't done any C programming since then). But even though I'm not writing C directly, knowing how things work at a lower level has helped me understand all sorts of concepts in higher-level languages.

[1] http://cs50.tv

I don't loathe C nearly as much as C++. I understand that learning about low level is useful. Pointers are indeed a very important concept (though not so difficult once you understand that variables themselves are already a form of indirection). So is Manual memory management.

But if you are tempted to put C++ in the same basket, don't. C++ is a very very different language from C. Sure, it looks like it just has some additional features here and there, but those features make C++ practices very different from C ones.

I've read way too many people online suggest Project Euler as a way to "teach yourself programming". This is almost a sadistic suggestion. Project Euler is way too hard to teach yourself programming.

Evidence: nearly 180,000 people have solved problem #1. Only 76,000 have solved problem #10. Losing 60% of your audience by problem #10 is pretty bad question design for learning.

Anecdote: I've been programming for over 25 years. I was near the top of my class in a top-10 computer science university program. And I've solved 30 Project Euler problems in the last 3-1/2 years. But I still can't solve problem # 18. I've worked on it off and on since December 2007. Only recently, quite by accident I learned that it is probably a "dynamic programming" question, which is something I've never done before.

Never lose confidence in your coding ability because Project Euler stumps you now and again.

"David Gries recalled a time when he was lecturing with Dijkstra in the audience when he read "x=a" as "x equals a". This is of course a short hand for "assign a to x" and not a statement that x really is logically equal to a. From the back of the room Dijkstra shouted "becomes" and the room fell silent. Gries thought about it for a moment and said, "Thank you Edsger, for correcting me. If I make the same mistake again, stop me." Everything went fine until twenty minutes later he made the same mistake when the shout from the back of the room came again - "becomes". Gries once again thanked Dijkstra and claims to never have made the mistake again."

The first language I was taught was Object Pascal, and the assignment operator ":=" was pronounced "becomes equal to".

So, between three of us, we saw that simple line of code in three different ways. I wonder what other people see in more complicated code.

Not correctly understanding assignment is one of the biggest failure points for beginners learning programming. The '=' operator has little relation to '=' in any form the students have encountered before. It means something like "evaluate the shit on the right hand side, and store it in a variable with the name on the left hand side". That meaning is not really even close to the meaning of the english word "equals".

You can personally call it whatever you want, but if you want to be a good teacher to beginning programmers, then you don't call both = and == the same thing. "x gets 5", "set x to 5" or "x is now 5" are all acceptable, but "x equals 5" is not acceptable for teaching purposes.

  x <- 5;

I agree with your main point, though, that failure to understand assignment is one of the biggest failure points for beginners. What programmers need to be able to do is trace through the execution of a program in their head, thinking "once this statement executes, the state will change in so-and-such way."

If I had to teach introductory programming again, I would focus more on having people fiddle around with variables in a REPL, just to get a feel for that sort of thing.

The only thing I don't like about video is that I can't do search for it. Sometime they show you a piece of code and you don't remember exactly where, you can't easily go to that part of the video.

It seems Khan Academy has a great model for instruction and Code Academy has a great model for practice; it seems reasonable to posit that a combination of the two would be superior to either one on their own.

I would further posit that the blend between the two models of instruction and practice would be something that might be tailored to the learning style of each individual student. Some people learn well by understanding before doing, others by doing first then understanding.