Before everyone gets too excited about the Digibyte idea, look at the followup tweet:
The idea is an algorithm that recreates itself every 10 days. FGPA's will be the ideal hardware to mine with. Every 10 days miners will get to "reprogram" and rediscover the most efficient settings to mine the #DigiByte #blockchain with.
Its not quite that simple when you understand how an FPGA works on the inside and is programmed; and most importantly how its internal configuration is made as to time each portion of its config to be in line and ready for each others sections to have done their work.
Prime example: look how long it took to get an xevan bitstream to come to pass; if you do the legwork and research that algo and its POW scheme; you can make some relatively minor edits to the algo to completely disrupt the timings and require better or more hardware.
With FPGA's being so complex (many aspects to getting them to do things properly), there is a lot people don't quite understand about them and how they can be effective or easily made obsolete...
If a hardware hashing device was based around a 10Mhz oven controlled crystal oscillator; and the algorithm has a time factor built in that can be related/based on this; it could easily become a standard for limiting hashing speeds; and could easily be re-verified as to be a valid result by the network using that time-base schism. It is of-course more involved than just a simple single function; but if worked properly; the time function can be integrated at steps throughout the process of calculating.
Example: Calculate for a result; and during each mathematical operation; do a math operation on the intermediate value/variable with the timer's current value as a salt; do the next step; salt, etc. All salt is based on the timer.
This way; you are forced to conform to the 10Mhz oscillator's salt, and it will disrupt your ability to hash faster as your result won't match someone else that is running the correct timebase, when you are accelerated and hashing faster.