To be honest i wonder if a good captcha can be solved automatically, i have seen this video about XRB, but the weakness was the unlimited number of try. I'm not sure a bot could solve a capcha at the first time, especially if it's a dynamic one (Once with image, once song, once number) for each fail you wait 1 hour.
However i have to say i really like the approach you propose. It's a very good idea.
There are two ways to automatically solve captchas: using OCR (for which complexity matters) and human solved (for which complexity is irrelevant).
The human solved ones make use of an API which distributes the captchas to people all over the world, who solve them manually and are paid for that. These are very cheap because they are solved by people living in places where that money is good money at the end of the month.
Potential errors are not relevant because the page is refreshed and you simply need to a good solving rate. Considering that this can happen continuously, a simple 50% rate is great, although it's much higher for the captchas used by RaiBlocks.
These automatic captcha solvers can also be run concurrently, therefore increasing the yield. It really is no different than a small piece of software generating proofs of work.
In fact, we might even be able to limit the generation of proofs of work by serializing them, by using a counter instead of a timestamp, and each proof of work would have to use the previous proof of work.
This would have to be combined with a registration process for the addresses that would be allowed to participate in the free distribution; for example, between dates X and Y, the users registered on this forum could send you their addresses (one per user); maybe also require a minimum number of posts.
The following would be a possible way to do it:
The process is similar to mining, but very limited in time (and therefore energy) and without the immense complexity of implementing mining into the cryptocurrency itself.
Make a small desktop application which generates a proof of work by using a memory-hard hash algorithm, like Argon2d. The goal is to make CPUs as efficient or more efficient than videocards and ASICs.
Each hash operation should take about 1 minute on an average desktop computer, and use about 2...4 GB of RAM. The chosen time is short because each verification of a proof of work takes the same amount of time.
Every 10 minutes, the application should save the last computed hash in order to minimize the possible loss caused by computer restarts and power failures.
The application would generate a proof of work, starting with the following information:
* A (random) seed which is generated (by the developers of the cryptocurrency) immediately before the final build is made and the application is publicly release. This will prevent the precomputation of the proofs of work.
* A timestamp.
* The user name from the Bitcointalk forum. This has to be registered with the developers of the cryptocurrency. Unregistered users would not receive free cryptocurrency.
* An address for the account which will receive the cryptocurrency.
* Counter which is incremented at every hash; this starts from 1.
The proof of work would consist in a single hash which has as many 0 bits as possible at the beginning of the hash. The number of 0 bits can vary so that the users can decide how many bits they want to generate (because they are rewarded proportionally).
At a specified date X, publicly release the application.
Then allow the users from the Bitcointalk forum to register their user names (and 1 cryptocurrency address per user), with the developers of the cryptocurrency. Unregistered users would not receive free cryptocurrency. Maybe also require a (small) minimum number of posts per user; the posts have to make sense, not simply be noise; this is to minimize the automated registration of users for the purpose of getting free cryptocurrency.
The application can be run for any amount of time, stopped, started again, and so on, since the computed hashes are saved periodically. The proof of work is the last saved hash. The number of 0 bits from the beginning of the proof of work is the "strength" of the proof of work.
At a specified date Y, for example 30 days after date X, the proof of work generated by the application has to be submitted to the developers of the cryptocurrency, by each user.
Each proof of work has to be validated by the developers of the cryptocurrency.
For each valid proof of work, the developers of the cryptocurrency would give to the registered user an amount of cryptocurrency proportional with 2 raised to the power "strength".
The amount of cryptocurrency given to a user should be capped in order to limit how much cryptocurrency a user could get by using a computer farm or a supercomputer.
