jherritz

Using the patterns on their own is a bit of an iterative process, but I'm sure you can imagine how it's a substrate on which all kinds of interesting automation of pattern learning could be built.

Interesting that you would bring up the concept of almost natural language. Barrage is actually built on top of Avail which is sincerely the most expressive language I've ever seen. Check it out: https://www.availlang.org/

It looks like there was a nice discussion not long ago on Hacker News about Snap as a programming language.

https://news.ycombinator.com/item?id=9591460

Do you have any thoughts about how you would implement dataflow programming in Snap? I know I'd be excited to see Snap hooked into some of the really interesting streaming data work we're doing at MIOsoft (beyond just the storage of projects).

I'm confident that is still an open question. That said, the creator of Snap would probably say there is no need to transition - ever. All Snap code can be viewed and edited as text. It's not an either/or decision, in other words.

You make some very valid points. There is an underpowered feedback loop here that would take years to amplify naturally. That's why it's important to have programs like BJC right now. It helps tilt the scales without relying on changes to the composition of the teaching field.

Clustering is a form of unsupervised machine learning. Many people are surprised to learn this.

http://en.m.wikipedia.org/wiki/Unsupervised_learning

Great idea! Synthesizing additional signals to obtain more accurate results is one of the core concepts of context platforms. Pulling in extra data from other maps is something I'd seriously consider if the simplicity of the demo use case wasn't the primary concern.