1) You buy old keys that at the same time in the blockchain history had 15% of all coins. Maybe some old exchange keys (BTC-E? ) will do the trick. (That is the "bribing attacker" Vitalik Buterin has described in his PoS analysis.)
2) You buy 15% of the supply and send them to your wallet, send them back to the exchange, privately mine a double-spend attack chain, and sell the coins again, and then you publish the chain for an attack.

In my proposal, none of the blocks or chains in the superblockchain are thrown away so none of the work is wasted.

Miners will create duplicate superblocks (not subblocks) because they are working competitively, not because the difficulty is low. The blockchains in the subblocks aren't forks because they aren't linked. They exist in parallel rather than in series. You can think of them as different but equally valid versions of the same transaction history

Each level 1 superblock, created by a level 2 miner, is just a short blockchain- a chain of regular blocks.

2nd level mining
Level 1 blocks are ordered and processed into blockchains to create level 1 superblocks and duplicate transactions are deleted so that only one instance can be spent. The hash is a little more difficult and the mining reward is a little better.

Well you're wrong. All mining schemes throw away redundant work (e.g. orphaned blocks). The higher the transaction volume, the more work has to be done and the more redundant work has to be thrown out. The numbers increase but the proportions say the same. None of it has a negative impact on security.

DO you guys ever think that bitcoin will do proof of stake? Just wanted to get some peoples insights on this.

coin been adjustment do huge diff  with asic's end now i try to fix it , but cannot find any info how do this.

if for some reason even 30% hash rate off any coin go away the transaction stop. Cataclysm, war with china etc... Quantum computer from NSA runed for 5 second end leave network.
This problem exist , it's problem not crypto algorithm - it's problem diff algorithm !

Ok use new algoritm how you say ? but if miner's left coin ? what to do ? You dont know ? whay you write on this topic ? Test net hv function for drop diff whay it not secure  end use ?

An increase in duplicate work but not forking. A cooperative mining scheme, like a mining pool, would eliminate duplicate work and be more energy efficient but it would also give a majority attacker total control over the mining process. Competitive mining is more secure and redundant work is unavoidable. A lower hash difficulty would be less secure in a single-level mining scheme but not in a multilevel mining scheme where the additional levels of mining work to secure the first.

any algorithm can be killed when for them create asic. It's not reason to mow to other algorithm need fix problem. If it's not fixed - even bitcoin can be killed with this.

So for example, you have one node that mines blocks as per usual, but then also have a reward for non-hashpower dependent tasks, such as confirming blocks? The logic behind this is that these rewards would incentivize people who dont have high hashpower to help improve security by vastly increasing the number of nodes that can confirm blocks.

right now, im thinking about designing an altcoin that uses both PoW to mine and then uses something like PoS, where everytime a block gets enough confirmations to be added to the blockchain, one of the nodes that confirmed it is picked at random and given a small reward (although nodes with higher stake would have better chance of winning). maybe it should be called "proof of confirmation"? People mining this coin would get to choose between PoW, which requires hashpower, or PoC, which doesnt

are there any problems with this idea?

Miners will create duplicate superblocks (not subblocks) because they are working competitively, not because the difficulty is low. The blockchains in the subblocks aren't forks because they aren't linked. They exist in parallel rather than in series. You can think of them as different but equally valid versions of the same transaction history. In any case, the elimination of duplicate transactions is part of the mining process at every level.

Each level 1 superblock, created by a level 2 miner, is just a short blockchain- a chain of regular blocks.
Each level 2 superblock, created by a level 3 miner, is a chain of level 1 superblocks and is therefore a blockchain of blockchains.
The subblocks of each superblock are different blockchains but they are not forks.

So the more dedicated miners are the common miners we know yet?

What is the difference than to 0-conf TX ?

How does the wallet=miner gets all the required data like last block, utxo,...? Only SPV style?

Forking doesn't happen until the last and most difficult level of mining, when the final level of superblocks are competing for inclusion on the superblockchain. Multiple chains of similar length may be competing and Bitcoin can handle it the same way it does now- by mining on the longest chain.

Well then you didn't sort them right, not my fault.

You ask excellent questions though, I love your face. 

It's not possible to avoid duplicates in your example but any two of the three blocks will include all three transactions so I can exclude one of the three blocks and quarantine the second instance of any duplicate transaction in a table entry that points back to the first instance of that transaction.

For instance, if I exclude block 2, I get duplicate A transactions in blocks 1 and 3 so, I quarantine the 2nd transaction of block 3.
([b3, t2] -> [b1, t1])

Not exclude, quarantine. Starting with the 2nd level, all but one instance of every transaction in a superblock will have to be quarantined so that only one instance can be spent.

Part of each stage of mining is the exclusion of elements that contain the same transaction data.