10kbp reads would also be very useful for de novo assembly of larger genomes, particularly in repetitive regions. One company (PacBio) currently claims reads this long, but the error rate is around 15%.
Science Fiction:
The article mentions cells that can "self lase" may eventually be developed. If they are, I wonder how well they'd work for neural interface. On the brain side, these cells could be used for output, with photo-diodes accepting computer input. For data transmission in the reverse direction, electronic lasers could be used to agonize photo-receptors in the brain.