[pufuwozu]

I worked on Precog. Here's my explanation:

It's an analytics platform for big data. Basically, you upload lots of JSON, use the custom language (http://quirrel-lang.org/) for analytical queries and the platform will handle efficient storage and querying of your data.

There are also has extras like a library for charting and reporting query responses using JavaScript and an IDE for writing Quirrel and uploading data.

[alexatkeplar]

Thanks for the explanation. What do you actually upload the JSON into - is it a proprietary NoSQL technology, something built on top of Akka, a wrapper around HDFS or something else?

[JulianMorrison]

How does this compare to rethinkdb (sharding, map-reduce, indexes, queries and joins on JSON data)?

[ddorian43]

columnar storage, compression etc

[JulianMorrison]

Interesting, so precog breaks JSON objects into columns dot-named by their path. I wonder how rethink does it, internally?

[joeblau]

How is this different/better than using something like Elastic Search? Does it handle real time upload/searching or do you have to bulk upload, process the data and then run your searches?

[pufuwozu]

No processing necessary. Just upload JSON.

The Quirrel language is much more high level than the API for ElasticSearch. It comes with things like machine learning (e.g. clustering and linear regression):

http://precog.com/developers/how-tos/machine-learning/

Even without libraries like machine learning, you can write somewhat readable queries which do some impressive things:

http://precog.com/developers/how-tos/common-analytics/

[joeblau]

Great, now I'm gonna be up for another three hours! Thanks for your feedback.