Thread by @Yazid

Before the release of my 'Automatism' project on @artblocks_io this Friday, I wanted to share some stories of the origin, background, and developments of the algorithm behind it, an algorithm which I came to call "Sharp-E"!
🧵👇

https://twitter.com/Yazid/status/1505189099732303886

As far as I can remember, this might be the actual tweet where I got the inspiration to explore an algo that does what Sharp-E does. I was fascinated by how artists like @shidahysf drew things seemingly out of nowhere to fill up an entire space.

https://twitter.com/Yazid/status/1431916904700993538?s=20&t=86rI_iN-GuqvyJg8hhrgrw

("Sharp-E" = Thick strokes like a Sharpie + autonomous and delightful like Wall-E)

I started out with a humble circle packing algorithm

https://twitter.com/Yazid/status/1440530700285845504?s=20&t=86rI_iN-GuqvyJg8hhrgrw

then quickly went from non-colliding individual circles to non-colliding strokes made up of many circles drawn closely together. I coupled this with a classic random walk algorithm which I had enjoyed playing with at the time.

https://twitter.com/Yazid/status/1440591788155170816?s=20&t=86rI_iN-GuqvyJg8hhrgrw

At this point I was using noise to determine the curves of the strokes in-flight, so parameterizing those values plus the maximum lengths of the strokes (how many circles a stroke was made up of) allowed me to start controlling how they looked and behaved.

https://twitter.com/Yazid/status/1440676713428500495?s=20&t=86rI_iN-GuqvyJg8hhrgrw

Playing with the starting angles of the strokes also ended up becoming quite a defining way for to me to achieve more variations in the outputs.

https://twitter.com/Yazid/status/1440851021446537226?s=20&t=86rI_iN-GuqvyJg8hhrgrw

As I continued tweaking and just plain having fun with generating random outputs, I started to enjoy seeing things appear in the outputs which resembled symbols or forms, like faces, letters, words,

https://twitter.com/Yazid/status/1442429122324566022?s=20&t=86rI_iN-GuqvyJg8hhrgrw

or inevitably, dicks. 😂

This was a more hardcoded approach I did for fun which obviously didn't make it into the final version, but doing this did teach me a bit about parameterizing certain behaviors, such as dramatic mid-stroke angle changes.

https://twitter.com/Yazid/status/1442498601687552003?s=20&t=86rI_iN-GuqvyJg8hhrgrw

These concepts of achieving a hand-doodled aesthetic, plus that of personal interpretations of meaning derived from the outputs of this "mindless" exercise of creativity, became quite central themes to the project.

https://twitter.com/Yazid/status/1442143322613583881?s=20&t=dNydLk2aVA4MgrxoyqOCjg

Credit for the name 'Automatism', which I felt fit perfectly with those themes, goes to a good friend of mine in this space, a confidant, and a long-time supporter of my art. If I am the father of Sharp-E, he would undoubtedly be it's loving uncle.
(Thank you @goartrover!)

In the beginning I hadn't set out to create Sharp-E as an algorithm to release as a long-from generative project. When I had applied to AB earlier in July I actually submitted an early version of what became 'Hashed Cities' as my proposal.

Initially I thought it wouldn't have enough variety. But by the time my turn in the queue came up, I had had a lot of fun with the algo that I got more comfortable with the idea. Early feedback about its diversity also gave me some much-needed assurance to take it forward.

(Thx to @ippsketch who demystified the AB process ahead of time, @bendotk @aaronpenne @piterpasma @amygoodchild @j2rgb @ioan1x @JosVromans @DataNeel @emilyxxie @rustysniper1 @williamapan @NadiehBremer + others I may have missed who offered feedback & encouragement!🖤 #genartclub)

From that point I continued iterating, and I also started testing out larger sets of outputs (100s at a time). But I quickly ended up getting into a bit of an infinite loop, endlessly tweaking and iterating but not feeling like I was getting anywhere.

https://twitter.com/Yazid/status/1448874041461264391?s=20&t=86rI_iN-GuqvyJg8hhrgrw

In fact I remember reaching a point where hated almost all outputs 😅

(This project was by far the most mentally-challenging art project I've ever undertaken)

I had also reached some performance limitations in Sharp-E's collision detection, as I had just been using simple vector distances then, which compared every new point to all existing points. I did start implementing a quad-tree to solve this problem, but I procrastinated. A lot.

I decided to take a looong break from it in November/December, shifting my efforts onto other projects instead, and I ended up releasing my first few long-form projects on @fx_hash_ which had just launched then.

https://twitter.com/Yazid/status/1459502649313619968?s=20&t=86rI_iN-GuqvyJg8hhrgrw

A breakthrough came around mid-January, when there was a Genuary prompt on "Packing", and it got me to revisit my efforts.

https://twitter.com/Yazid/status/1481072284559638529?s=20&t=86rI_iN-GuqvyJg8hhrgrw

I also had a chat with @piterpasma then who quickly made me spot some oddities in my code. His other big contribution was suggesting to explore additive sine waves for my curves. I took it up and ended up loving it.
(Thank you Piter!)

The last major change to the algorithm happened post-AB curation board review. I sought out feedback and learnt that they would've liked to see the algo lean more into the "drawn by hand" idea. Up to that point I had mainly kept the strokes fairly clean.

https://twitter.com/Yazid/status/1492172582920998915?s=20&t=86rI_iN-GuqvyJg8hhrgrw

I worked at it throughout an entire weekend and added a feature which introduced some subtle textures to the strokes, simulating different pressure points, angle differences, fraying in the marker tips, that sort of thing.

https://twitter.com/Yazid/status/1494859967584772098?s=20&t=86rI_iN-GuqvyJg8hhrgrw

Then I took it further and turned the dial up to 11.

https://twitter.com/Yazid/status/1496281773801713671?s=20&t=86rI_iN-GuqvyJg8hhrgrw

One final thing I added was a slight ease-in in the drawing animation of the strokes. That, plus the periodic auto-reassembly of the strokes, to potentially facilitate re-viewing if displayed on digital frames. This I have @ArtOnBlockchain to thank for.
(Thank you Erick!! 🤝)

Some of these last few outputs I quote-tweeted were early iterations which got developed further for the final version. You can see the eventuality of the algorithm in the Ropsten sample outputs, which I hope to speak more of soon.

For now, thank you for tuning in this far! ✌️🖤