No, the best way to learn a language is comprehensible input. Every other language acquisition method is bootstrapping that eventually needs to segue into actually using your target language to read or listen to things, unaided.
What these bootstrapping exercises are doing is not unlike, say, what early expert systems or Cyc did with AI. They aren't so much building an understanding of language as much as they're handing you a bunch of logical rules to parse out into sentence constructions. The problem is, that's not how human language actually works. In fact, it's not even how humans use programming languages, even though those do have formal specifications.
If you want an "open source Rosetta Stone" what you want is Anki and a flashcard deck for it. But even then, that's limited to vocabulary memorization, which is just bootstrapping. Personally, if you wanted to build a good language acquisition app, you almost certainly would want to have some kind of large language model in there powering it.
What these bootstrapping exercises are doing is not unlike, say, what early expert systems or Cyc did with AI. They aren't so much building an understanding of language as much as they're handing you a bunch of logical rules to parse out into sentence constructions. The problem is, that's not how human language actually works. In fact, it's not even how humans use programming languages, even though those do have formal specifications.
If you want an "open source Rosetta Stone" what you want is Anki and a flashcard deck for it. But even then, that's limited to vocabulary memorization, which is just bootstrapping. Personally, if you wanted to build a good language acquisition app, you almost certainly would want to have some kind of large language model in there powering it.