The foundations of information embedding were laid more than twenty years ago with early information-theoretic work on the problem of channel coding with side information at the encoder. One of those early papers was Costa's whimsically titled "Writing on Dirty Paper". In recent years, four key advances have made possible the development of practical information embedding systems. First is the recognition that the information embedding problem can be cast as a problem of "writing on dirty paper". Second is the development of practical code constructions in the form of distortion-compensated quantization index modulation (DC-QIM). Third is the development of capacity-approaching turbo codes with very low complexity iterative decoding algorithms. And fourth is the maturation of VLSI technology and the associated development tools. In this project, all four of these advances are exploited together in the design of the world's first real-time, capacity-approaching information embedding system: the "dirty paper" coder/decoder (codec). This system transparently embeds a bitstream at 6 Mb/s into an arbitrary analog NTSC television signal so that it can be reliably transmitted through, for example, any FCC-compliant cable network. Such systems could be used to add roughly 300Mb/s additional downstream capacity to cable infrastructure without requiring plant upgrades or affecting existing services. The system operates within a couple of dB of capacity, and the reference design is complete with analog front end, 400k gate FPGA codec, full DSP-based software equalization, and an embedded linux controller. Other emerging applications of this technology include multi-antenna downlink transmission in wireless systems.