Done with the Language Transfer guidebook. So it's somewhat similar to Michel Thomas in that it's Q and A and you are asked to produce sentences pretty much from the beginning and it's mostly about grammar (he even got sued by the MT IP owners because of a patent for the trivial idea of using the pause button in an audio course!). The sentences start very short and get more and more elaborate. You are nudged towards guessing words following shifting patterns between related languages, with MT being less transparent about it than LT. MT is more about giving you a feeling of what's right and it's not so clear to the student just how much you are nudged. This also means that with MT you are not quite as competent as you feel you are, but it's still a great way to get into a language.
The LT guidebook is not as enthusiastic as MT about using shitty puns to remember vocab. It makes more use of real etymology (history of a word's usage and how it mutates over time and with location changes) for making connections. That's not to say that you never should use puns and made-up etymologies, but memory-hooks don't come for free. They require mental energy. An example of a shitty memory-hook is imagining two lovers in the rain for remembering that Spanish llover means to rain. Here you are putting up with sticking half a sentence in your noggin just to remember one word (and it isn't clear to me how you would remember it's a verb since lovers and rain are nouns after all). So for a story to be worth it one should be able to combo into several things from there, and the chance that this works with stuff you make up spontaneously as a newbie is low. But real etymology, real story, tends to work in patterns. I think it's still worthwhile to come up with silly stories for letter strings in the target language if they appear in many obviously related words or if you can make them relate through the story.
It isn't hard to remember stuff like spelling new words if you have a good feeling for how the language usually works, so you have to only remember the expectation-defying bits. Getting the student to that point isn't done well usually. Languages tend to have some verbs with high frequency usage (like to say or to go) that are conjugated in ways completely different from how it usually works. So not to poison the intuition-building process, it makes sense to only let the exceptions drip in slowly, and show more of the normal way. One should also avoid cryptic grammar terms if possible, as LT does. LT is really good and I'm surprised it isn't more well-known.
I see a similarity between ways of teaching and ways of writing a newspaper article. When writing in classical newspaper style, you assume that your audience is busy and might quit reading after any paragraph. So, you put the gist in at the start and then add details and after that mini-details for the macro-details. You can't do things like arguing strongly for one position and then put a twist in the end. But what if your readers actually go all the way to the end? What you write is going to be more samey and boring than it has to be. Gabriel Wyner makes the point that learning vocabulary by thematic sets like colors is boring. It's more easy to remember random vocab lists where each word is more likely to stick out as special in its own way, and better than that is using sound-shift patterns specific to the base-target language combination or sets that recycle word fragments. Likewise it is boring to learn grammar in a way that drills down a specific section (like all present conjugations) before anything else, a point that the LT guy makes. So, why is language usually taught the boring way? Because it makes it easier for the teacher to track where the students are and it makes it easy to communicate this to people who don't know anything about the language (and it makes it easier to replace the teacher). If you do a third of a regular language course, you can easily summarize what you know. The topics in LT are so densely weaved together that the lessons are just titled by number and the revelation of the grammar map is done with many jumps revealing small spots.