Disegnatori (Italian for 'draftsmen') was released on fx(hash) on November 29th 2021 as generative token 1641 with an edition number of 250. The last iteration was minted on December 3rd.

Each disegnatore learns a different portion of a city map, looking at it stroke by stroke and creating a probabilistic Markov model of the sequences of relative angles and distances, tracing the outline of each building in that portion of the map. This way, each disegnatore has a different internal understanding of the city.

It then starts from the middle of the canvas and wanders out, stroke by stroke, probability by probability, trying not to intersect any previously drawn outlines and avoiding the edges of the canvas until a predetermined number of edges have been drawn.

The Markov chain drawing model is created during execution from a text file representation of a city map.

The map that the disegnatori learn from consists of strings representing polygons of building footprints in a format designed particularly for this project. The strings contain pairs of ASCII characters describing relative heading (ASCII code 32-79 for 0 to 360 degrees in 7.5 degree increments) and distance (ASCII codes >= 80, as the closest integer of the square root of the distance in meters), respectively, for each line segment.

For example, the strings W+U+Q+QDRDROS+R U SOR,S,UCSDT+U and T+SDTGT,T Q+T"S#S"V+T S describe the following polygons:

The Markov model of the map can be used to generate similar strings that can be converted to heading/distance pairs and then into points describing the open or closed polygons.

The ones below are described by the generated strings RER+T-RBT U,U,RDW-S+VDWDS,T,TCSCT, S,U,S+QDRCTDR+S+SDSDS,R+SCQ,SDT, S)V,T,SCSOR ROR RCR SOR+VOW0V W and UOROTCSDTCTDTCQ,SDQCQ+Q Q+Z,QCUEUCR,S+QDS,SBT QDS+T,T. As the model learns the polygons in the map with a fixed-length window, the generated strings can yield polygons that are open and/or self-intersecting.