Finished first beta version (1.0) with capitals and some punctuation glyphs. I tried to keep most of the captial letters as uncomplicated as possible to balance the bulge in the lowercase.

Looked at Arial's rules and worked on proportions. Cleaned up smaller features to create a more coherent look in body texts. Still unsure about f, g, r, t, y and z. Starting to work on capitals now.

Started to implement individual feedback. Worked on a, c, f, j, t, v, w. Next step would be to open up letters like b, d, p, q.

Did all lowercases, now reducing the features again, too much going on. Want to focus on two or three special letters and keep the rest blended in. Fav letters: "a", "s", "u"

For now I generated the standard "adhesion" with Arial letters as input data. Always two letters as inputs: so for the "a" I used "a" + "d", for the "d" I used "d" + "h" and so forth. After the training of the network I generated 100 frames of each letter iteration, leaving me with eight animations of how the network tries to recreate the specific Arial letters. Now I looked at the individual exported frames and focused on the features, which seemed interesting to me and drew my first adhesion from there. These 800 frames will be the only use of machine learning in the process. I will not generate more letters or try to automate the machine’s output further. Conceptually it is important for me that, even though I used Arial as a base for the experiments, I’m starting from scratch, drawing all letters (by hand) without using the images as a direct reference. The process and intertwine of human-machine decision making is probably most important to me.

Still trying to figure out the main features, also working on the overall coherency as a non display font. But will focus on "adhesion" now first to finalize these letters and then go from there.

For now I'm just looking for interesting features, not necessarily a coherent appearance. The idea is to focus on one or two features and transfer them onto the rest of the typeface. I’m not sure if I should focus on Arial as a base, since I really like low contrast sans-serifs or also try as set with Times New Roman. Conceptually it is important for me that, even though I used Arial as a base for the experiments, I’m starting from scratch and drawing all letters by hand without using the images as a direct reference. The process and intertwine of human-machine decision making is probably most important to me.