This is a fantastic idea that would be really cool if it worked... but is there a cryptographer in the room that can verify that it actually does work? There's a dime-a-dozen "secure" communication methods that aren't really secure.
According to the github description: "To send a message via IronPigeon, a pair of endpoints must exist. Endpoints have both public and private components, containing the public and private cryptographic key pairs respectively. When party A shares its public endpoint with party B, party B can send party A messages. When two parties each create their own endpoints and exchange their public components, the two parties may communicate securely."
The crypto practice of using both public/private keys is good. It makes sure that the communication is based on a trust scheme where one person must trust the other however there may be ways to fake such things depending on how they implement the crypto.
The messages should be put through AES 256byte encryption or similar so there is no reading the message while it's in transit till it is recieved by the right address. Instead it's left up to the user how they encrypt their messages.
However I don't think I would trust it that much primarily the main thing that this is trying to sell is that it is like doing e-mail messages that are decentralized in circulation, have an expiration time limit, and the messages require the reciever to set up an endpoint to recieve messages.
It's a good start, but it needs better explanation about some of the encryption that is used.