It wasn't that clear to me from the video what was being offered by your service and what was being done with tmux or other CLI tools.

Is the benefit that you will spin up an EC2 instance from a config file in a repo and then me and my friend can work on it? Or is there more benefits to pairing with your service, like the collaborative editor?

A server is spun up and credentials given to all the invited users. Once a user logs in they can connect to a common tmux server and share a text-editor/cli/etc.

I'll save judgement for when I could try it, but why on earth do you need full access to my git hub account? I don't want to give you access to my private repo's unless I'm coding on them.

If I remember correctly, tmux has this feature where multiple people can connect to the same window (shared sessions?). Is this the feature of tmux being used? If so, what does the pair.io server do? Where is the tmux window hosted?

I can see this being useful if you do want to pair-code though I still see pair coding as clunky, especially since only one person will be editing the file at a time anyway.

that I don't understand either. You can do the same thing on your own machine, just add a user and tmux there, where you have the current code base up and running.

What is used for the instance configuration? What assets/scripts come from pair.io as opposed to what comes from my repo?

Project-specific configs can be committed to a repo to override the default configuration.

The GitHub integration is a nice touch!

So is the message you get if you try authenticating with GitHub without being in the private alpha.

Looks like there's some cool technology here, but I'm not sure why it has to spin up a new server.

Some projects are self-contained, and for those you don't need any supporting infrastructure outside the runtime of your language. But once you start dealing with _systems_ where it's very particular about the specific version/configuration of the database or message queueing system, it's much less error-prone to have that automated. For that you need control of a whole machine, virtual or not.

I wrote a blog post about a similar approach applied to local VMs: http://technomancy.us/150

So you can't resume an instance? I only ask because some of my projects take a long time to import the initial seed data.

The intro video could be improved, it dragged on for a little too long, but it seems cool!

This is pretty awesome. Especially since we use a digital team methodology.

Am I missing something here ? We do this using screen/screen -x

His hello world function is missing []