...
Dero should make a algorithm that randomly generates a new algorithm each block. It sounds crazy but I have faith in the Captain's abillity to code. He was able to re-write an entire protocol in a new language in 3 months. This guy is an expert in cryptography and algorithms so creating something totally unique isn't a question of can, its a question of will or won't.
...
Interesting idea that sounds somewhat like what RavenCoin and PigeonCoin do. They use 1 of 16 well known/established PoW algorithms on each block, with the algo chosen randomly for RVN (X16r) or shuffled for PGN (X16s). While this approach can theoretically be ASICed just like any other, it would require a massive undertaking to implement all 16 algos on one chip and therefore makes it supremely uneconomical.
Note, however, that one of the things I like about mining CryptoNight is it uses much less energy than, say, Ethash or Neoscrypt (to name 2 algos that do well on AMD cards). Conversely, one of the things I don't like about CN is that it sucks on Nvidia cards (a $1000 GTX 1080 can barely exceed a $150 RX 560).
While I do agree that DERO would be seen as a trailblazer for coming up with a truly ASIC-proof algo, I also agree with @cpmcgrat that this would be a major distraction versus pursuing smart/private contracts. Also it would invalidate all the work that CaptDero just did on rewriting DERO in golang, and I rather suspect that would be difficult for him to toss out. Hence I still think that either going with CryptoNight Heavy or following Monero's lead are the least-risky courses to choose from.