Ask HN: What are you working on and why is it cool?

I've been working on this full time, completely solo, for nearly 11 months now. It's lets you import any content to read, provides low friction translations so you can concentrate on enjoying the content, and has a spaced repetition system to learn words and phrases with flashcards.

It's getting some early traction now with 2300 users, 45 of whom are paying, and really good feedback.

There is no storage on the back-end, it just calls a little node library that I wrote and spits out the result.

The work to target other languages is underway, I've split out the AST walking to a separate repo, so it just requires writing a walker implementation for the target parser and you're done. The CoffeeScriptRedux walker is currently in development, with LiveScript planned for after that.

So I met a guy recently who is building something similar to the ComplexityBot idea you describe. It's called sidekick and looks pretty awesome from my initial play around with it.

* https://www.sidekickjs.com/

Adversary model makes sense. I think security/privacy folk like reading up on them for a given project they might be curious about, perhaps consider putting this on the site?

> A global passive adversary would be able to pinpoint the origin of posts if they had access to the wire-level data if posts were sent in cleartext, but all the connections are encrypted with TLS, so that's impossible.

How about timing correlation attacks, etc.? (Basically, all the attack vectors Tor people should be afraid of?) Do the posts have timestamps attached to them? If yes, wouldn't a GPA (or, for that matter, a local ISP) be able to correlate certain node activity with posts?

> If you have left locally flagging own posts as such enabled (the default setting), it will identify the posts you have created. But if you explicitly disabled that option, even that does not incriminate you: your node is just another node that might have been touched by that post.

This seems like a very nice feature / piece of design, cool.

> The reason I have that 'own post' flag is that it is the only way I can show or notify the user about the replies he / she receives.

Have you considered abstracting this to 'subscribe to this post/thread and be notified of any replies' functionality? (anyone could subscribe to posts, then (presumably with read access / however it works.))

> The only reason I have protections against that kind of threat also is the possibility of people in less fortunate countries (where bar for seizure is low) getting their computers seized because of their exercise of free speech.

As far as I'm aware, there are indeed instances when people get their equipment seized, and upon detection of sensitive (to the regime/power/$thing) data, bad things may happen. That's why Tor is very careful about what it writes to disk, etc.

Overall, thanks for your work, I'll try and follow the news and maybe take a look at the code. Good luck!

As for escaping special characters: I've run into that a lot, and I agree it is very annoying. God help you if you want to specify regexes on lines.

So for a lot of my YAML-ish applications, where YAML is used for user templating or configuration, I run everything through a really simple postprocessor (often by encapsulating non-digit lines with single quotes, and a few other tweaks) and that seems to work well for me. That way the user only has to type the bare minimum (no double quotes, no curly braces, no/fewer commas), and what they write "just works."