In 1948 Claude Shannon showed that information can be measured, that every source has a size no compressor can beat, and that every noisy channel has a speed below which messages can be sent with as few errors as anyone likes. Run on the page: surprise and entropy, mutual information and divergence; Huffman, arithmetic and dictionary coding; channels, capacity and the Gaussian limit; parity, Hamming, Reed–Solomon, Viterbi, LDPC and polar codes; and information in secrecy, statistics, learning, physics and life. Count the bits, build the codes, set the noise and watch a code close in on its bound — every entropy, rate and error computed by the page.
Every code, curve and number on these pages is computed by the page or cited with its source; your progress stays in this browser. Secrecy is taught as information theory: the ciphers shown are teaching ciphers, long broken, and nothing here is a way to protect real data or to read data that is not yours. Real systems — ZIP, JPEG, the compact disc, QR codes, Wi-Fi, mobile networks — are explained with the page's own teaching codes; no standard is implemented in full. Nothing here is medical, financial or engineering advice. The course endorses no product or company: systems and companies are named only for their place in the history of communication. Not affiliated with or endorsed by any organisation, standards body, publisher or author mentioned here.