My ideal world scenario is that I can use Codex like I have been using it, to exchange chunked datasets. What I have been missing the most:

1. Light clients (mobile, browser);
2. Being able to publish a dataset without a public IP;
3. Pin to a node;
4. Force remove;
5. Range headers would be cool for Podex.

Václav also says:

"... So going with altruistic mode, doing what Ben came up with (constellations) or similar things (like what I did with Qaku Cache) would bring users -> grow the network and awareness -> give us more time to make durability useful for more people ..."