Small fixes:

• Stopped the Duck Kami sticker from looking like a Bleeper when you put it on a letter.
  
• Fixed not being able to delete responses if your inbox was sorted by request.
  
• Changing your signature initial used to have a delay before being reflected locally.
  
• Clicking on the version on the title screen now takes you to the steam news page instead of the static game website.
  

That's all. If you're like me, you might have noticed an update for a game you care about and then clicked on the news only to be slightly disappointed that it's just bugfixes for bugs you didn't even notice.
Sorry. That's all that needed changing right now. But I CAN give you a little update on some behind the scenes dev stuff if you're just hungry for some Kind Words goings on.

Moderation is a big job for us and we take it seriously. We're always looking for ways to do it better and more efficiently.

You ever read those jokes where it's like: "We fed 10,000 movie scripts in to an AI and asked it to write Robocop 5!"

Well... we've actually got the sort of dataset with Kind Words that just might be right for that kind of AI. We've got around 4 million organized text entries and 200,000 moderation actions. Now, "feeding it to an AI and asking it to moderate" isn't entirely straight forward. There's a bit of data science and statistics to do. And the biggest catch is: I need to do it myself. We take our privacy policy very seriously. I'm never going to hand off our database to a third party, no matter how good they are at feeding it to machines.

So that's what I've been up to. Dusting off my python. Reading up on support-vector machines. Connecting PHP and Python in ways God never intended. I've got a LONG way to go, but I'm excited to say I did a first test in the production environment today and it has already flagged several questionable messages for review. I don't know if it'll ever get to the point where I trust it to make final decisions, but damn is it cool to be trying.

Cheers,
Ziba