thesis drafts → preface aka postscriptum
Key concepts: Rhizome, Deleuze, CCRU, Haustorial Web, the rise of LLMs and the new writing paradigm.
In one sentence: I aim to combine two approaches while writing this thesis, one is unstructured and freeform that this page is a part of, the other more academia-friendly that will exclude some chapters and have a rigid structure.
LLMs are good at structuring things -> as a writer i want to explore the opposite direction
Academia has standards and i see little value in trying to push them, so there will be two versions of this work, one printed on paper in linear fashion and another here online as a constantly growing and changing interconnected micro-database of relevant ideas. It might live on and grow beyond its original form, or it might naturally die like all things in life eventually do.
My master thesis, as per official requirements, consists of two parts: theory and practice. Every part, in turn, consists of two smaller parts: graded and non-graded. This page, for instance, is not graded, which means I will not include this text in the PDF file that I will formally submit to weißensee kunsthochschule as my thesis paper. I have marked every non-graded page with an asterisk (*) and approached them with more freedom than the graded part — some of them might be written in a less formal language and/or include external links and other media. These pages serve as a glue that will hopefully hold my ideas together no only in a straightforward linear way, but also rhizomatically
In 2020, I defended my bachelors thesis in cognitive science. The title was “The Effect of Intersensory Cooperation on Absolute and Differential Perception Thresholds”, which is a sophisticated way to say that I researched how vision affects sound and vice versa. In February, as I was still working at a local nightclub, I would invite random party goers to the technical room one by one, and burden them with a 15-minute test in which they would see flashes and hear beeps in a very repetitive manner, and then report how many flashes they saw and how many beeps they heard. Some of them were happier to participate than others, but I had to be persistent: I needed over a hundred samples for that kind of experiment and I hardly ever went anywhere else except that place and home. In March, the lockdown began and I got trapped at home like everyone else — that was perhaps the only setting that could make me fully focus on the paper and write it free from distractions for two months. Our generation was the only one who got to present and defend their theses over Zoom, and in the end it glitched so badly that the board, consisting mostly of people over 70 years of age, gave up and just sent us their questions in the chat, asking to respond by
I am writing my master thesis in 2025 — the year in which large language models like ChatGPT and Claude are already capable of producing
LLMs are very good at structuring things — their vanilla “let’s break it down” style of writing is still easily recognisable to the experienced eye (although they can alter that when explicitly asked). What they are not good in, though, is breaking patterns: a model cannot jump off a train of thought into the abyss and end up on another one.
In the following (or preceding) text I try to stay away from rigid structures, and embrace the chaos of my own thoughts whenever appropriate. To keep things relevant and coherent, I will
It feels to me like in this particular point in time, even though it might be non-evident (or plain wrong), there is little value in making text make sense in the conventional way. It will take us some time to adapt to this reality, and we will eventually change our criteria of evaluation for both human and machine-written text,