Notes

https://storage.googleapis.com/deepmind-media/Era-of-Experience%20/The%20Era%20of%20Experience%20Paper.pdf

This paper claims that the next big leap in AI won’t happen by giving models more human generated data. AI agents must learn through their own real world experiences instead of relying on human data. We already use RL (reinforcement learning). For example AlphaGo learned by playing millions of games, and coding agents can learn from whether tests pass. The idea isn’t new to reinforcement learning, but moving from AI that learns what humans know to AI that discovers what actually works. It also mentions that the big challenge is making that continuous experimentation safe in the real world.

Imagine an AI learning to grow cherries. Today, it might read thousands of gardening books and tell you how much water and sunlight cherries need. It will take the human knowledge of the physical world as the base fact. In the Era of Experience, the AI could grow cherries itself, try different conditions, observe the results, and discover better methods that nobody has written about. It’s a feedback mechanism involves real world experience.

Today’s RL often optimizes models using predefined tasks, reward models, preferences of human, or automatically verified answers. In the paper, for the proposed “era of experience”, AI learns continuously from the consequences of its own actions in an environment, potentially over months or years.

The authors defines the AI eras: 1. Era of Simulation: AI learns by trial and error in simulated environments. Examples: AlphaGo, AlphaZero, Atari. Very powerful, but mostly in narrowly defined problems with clear rewards. 2. Era of Human Data: AI learns from human-written text, code, demonstrations, and preferences. Examples: GPT-style language models. Broadly capable, but constrained by the information and assumptions available in human data. 3. Era of Experience (their prediction): AI interacts continuously with the real world, tests hypotheses, receives feedback, and learns strategies that humans may never have imagined.

A sketch chronology of dominant AI paradigms. The y-axis suggests the proportion of the field’s total effort and computation that is focused on RL.:

In the era of experience (direct quote from the paper)

• Agents will inhabit streams of experience, rather than short snippets of interaction.
• Their actions and observations will be richly grounded in the environment, rather than interacting via human dialogue alone.
• Their rewards will be grounded in their experience of the environment, rather than coming from human prejudgement.
• They will plan and/or reason about experience, rather than reasoning solely in human terms

To achieve this era the agents must interact with the physical world (direct quote):

To progress significantly further, a new source of data is required. This data must be generated in a way that continually improves as the agent becomes stronger; any static procedure for synthetically generating data will quickly become outstripped. This can be achieved by allowing agents to learn continually from their own experience, i.e., data that is generated by the agent interacting with its environment.

I saw this paper from a Youtube video by WelchLabs. There in the video, an interesting queston was asked as a thought experiment. Actually the experiment also coming from the paper itself.

This is my favorite LLM thought experiment. If you trained an LLM using a really old cutoff date for your training data, like 1000 AD, what would happen? If you asked your 1000 AD cutoff model what made metal shiny, it would probably reason in theistic terms (answer: The LORD made that way). If you went further back to some of humanity’s oldest texts, it would probably answer in terms of animism (answer: animal spirits inside metal). If your cutoff was around 1700, this LLM would likely explain metal shyness using Newtonian physics (answer: light reflecting on smooth surface), and a more modern model would answer in terms of quantum physics (answer: free electrons absorbing and re-emiting photons).

An interactive web based audio visual artwork exploring Rule 30.

I recently watched this amazing video about rule 30 showing the beauty of this from Eric Rowland: https://www.youtube.com/watch?v=HcGjCcLEgN4 There is a prize for rule 30 https://www.rule30prize.org/ and it makes this more exciting.

After seeing it I wanted to iterate some simulations using audio and visual elements.

Explore the project: https://hasantayyar.net/assets/rule30/

The project uses Web Audio API and there are modes:

• Tapestry: watch and hear the automaton.
• Hidden order: reveal repeating right-edge diagonals.
• One-bit echo: change one cell and compare the resulting futures.

Click Listen to enable sound. Space pauses, 1–3 switch views, M mutes, and F opens fullscreen.

The wave is a visual effect; cell values follow L XOR (C OR R). Finite observations do not prove randomness.

Inspired by Eric Rowland’s research.

I read about Zā’irja, a medieval “combinatorial text generator” device. I can’t accept most of the claims, but one part was making sense for me: a set of terms/words can produce questions I would not ask on my own.

I vibe coded a small web app from this idea. The following app combines a problem, a way to act, a constraint, a new point of view, and most importantly human goal.

Explore the project: https://hasantayyar.net/assets/possibility-engine/

I was reading “Scrambling T-R-U-T-H Rotating Letters as a Material Form of Thought” by David Link.

I stumbled upon that article while reading about Ramon Llull’s combinatorial machines.

From there jumped to the Zā’irja, another paper device/machine. Historians accept that Llull was heavily and directly inspired by the Ibn Khaldun’s Zā’irja**. Ramon Llull was living in Majorca and learned Arabic. He was exposed to Islamic science, astrology, and mechanics.

Zā’irja is also a paper machine, a paper computational machine, like Llull’s machine. It’s widely analyzed and demystified by Ibn Khaldun. The device originated in the Islamic West (Maghreb / Al-Andalus) around the late 12th century—roughly 200 years before Ibn Khaldun wrote about it. The Zā’irja was many years, like nowadays numerology, used by magic charlatans and misunderstood. Ibn Khaldun completed the first draft of his work/book Muqaddimah (The Introduction) in 1377 CE. And he mentioned about Zā’irja in his work widely. Although it’s mostly very poetic and not to the point. He wrote it while he he was in self-imposed isolation at the castle of Qal’at Ibn Salama, nowadays Algeria. https://en.wikipedia.org/wiki/Muqaddimah

Zā’irja operates like an analog computer or a systematic ideation tool designed to generate answers to complex questions.

How It Worked

1. Input: You input a clear question written in Arabic.
2. Breakdown: The letters of the question are converted into numbers using traditional letter-number systems (Abjad numerology).
3. Processing: The data is run through dynamic, concentric circles containing astronomical, cosmological, and mathematical rules.
4. Output: The machine rearranges the elements to produce a poetic or insightful answer.

I’ve been using Ghostty for a while. It is one of those tools I installed out of curiosity and just kept. It was created by Mitchell Hashimoto, co-founder of HashiCorp and the person (one of the people) behind Terraform, Vagrant, Vault, Consul and Nomad.

Ghostty is written primarily in Zig. The language itself is already focused on performance. It Ghostty to combine speed, modern features and a native UI without becoming complicated. Mitchell shared every step while developing the terminal. It supports GPU-accelerated rendering which I believe the most important feature, tabs, splits, ligatures, images and modern terminal protocols. A fast terminal matters more than it sounds: it sits in the feedback loop of almost everything we do, so tiny delays repeated thousands of times become noticeable. First users always express their excitement.

Once the terminal becomes fast, the bottleneck is often the shell setup itself. No flame war needed and zsh itself is not the problem. It is a shell, while Ghostty is the terminal emulator. What often becomes slow is a heavily customized zsh setup (Oh My Zsh, plugins, completion systems, hooks, etc) which can erase much of the snappy feeling you gained from a fast terminal. ; Even Oh My Zsh warns that too many plugins slow shell startup.

Mitchell has since co-founded Superlogical, where he is building a new terminal multiplexer on top of libghostty and continuing to rethink terminal-based developer workflows. Ghostty feels less like just another terminal and more like an attempt to modernize one of the oldest interfaces we still use every day. The interfaces which started to matter more nowadays. Because terminal interface is actually the most flexible and personlized interface so far.

See his latest pre-demo but very nice sneak peak video “A quick look into remote persistent sessions with Superlogical!” : https://x.com/mitchellh/status/2097424868203758046

References: Ghostty docs · Features · Source code · Hashimoto · Superlogical

I was born and raised in Imrahor, a valley neighborhood in Ankara that lived inside strange conflicted feelings. This entire neighborhood was abandoned, self built small houses, green, autonomous, forgotten yet so close to the historically bureaucratic and aristocratic center of the capital. A patch of wild area where wild birds stop by when I was a child. The neglect within walking distance of power.

For decades it existed in that liminal state: half-inhabited, half-erased from official memory. Also for decades this area was considered to be the lungs of the city and had great potential. Then one day it was demolished to be replaced by tall buildings. Not gradually gentrified, but cleared entirely. The valley opened for whatever the city’s next transformation demands.

Back in 2023 when the razoring of the neighborhood happening, I went back with a camera to freeze what remained: an entire neighborhood giving away its last breath, surrendering the last traces of memory still held in crumbling weak walls and scattered rubble. These photographs are not documentation of progress. They are a record of what urban transformation costs when the people who carry a place’s memory are not the ones who decide its fate. A record of the power of the state razoring what people build with their own hands.

I have the new photos as well. But that is too ugly to show. Simply tall buildings, typical Turkish urbanization.

The new city already waits on the edge of the hill, creeping in to greens. Below it, the valley floor is stripped back to earth, stone, and whatever could not be carried away.

Houses that once held generations reduced to weight and volume, waiting to be removed.

Photographs taken in Imrahor, Ankara - March 29, 2023 by Hasan Tayyar Besik

Previously: EIL August ‘26

August became an unexpectedly least physically active month. I broke my leg and had to slow down for at least a month, which left plenty of time for documentaries and my old archives.

Famous painter from a small town: Ivan Rabuzin

During a karate camp in Croatia (end of September), between trainings, I visited the home of Ivan Rabuzin, the internationally known painter from a village near Novi Marof. I love it there around Novi Marof.

Ivan Rabuzin is long gone and his son-in-law is working to preserve his house as a museum.

Rabuzin began as a dishwasher, learned carpentry, then turned to painting. A quick summary I learned from his son in law. I learned too many other things. They try to make the house a museum. He showed Ivan’s one of the early works was reportedly used as a chicken feeder before his work became famous. His paintings later found an audience as far away as Japan, where one was used as a theatre curtain.

Rodna kuća - the birth house - and Stara Klet, the old cellar, before the museum took shape.

The house is as interesting as the paintings. Rabuzin helped design it and made its furniture himself: chairs, cradles, benches, too many plates and ceramicware, and a huge dining table. It reminded me of the distinctive private architectural projects of old Ankara. The house also has spectacular but humble details to show the craftsmanship.

I started loving his unique style. Check out his work https://marilyncvitanic.substack.com/p/the-art-and-inspiration-of-ivan-rabuzin

Later in the area we have realized that there are his art work on top of the plastic water bottles as well. Apparently the water company made a deal during the socialist years for one time purchase of the usage rights from Rabuzin and still using that contract.

At the garden of the house there was a metal installation looks like a very big burnt match stick.

Barkley Marathon

I started watching documentaries a lot while recovering after my broken leg surgery. Among others this one was interesting. A docu about Barkley Marathon. I love the obsessed amateur spirit, ambitious community and hard trail and along all the fun historical elements that built the trail. The founder is himself has a great smile during the entire documentary. It was a feel good documentary for a person with broken leg in recovery :)

https://www.mattmahoney.net/barkley/

From the website:

The Barkley is considered one of the toughest “100 mile” races in the world. It has (67000 feet) 20 km of climb (and 20 km of descent), more than any other 100 mile race, more than the 33,000 ft. of climb at Hardrock, and more than the 45,000 ft. at Nolan’s 14. Since the race began in 1986, less than 2% of runners have finished within the 60 hour cutoff. Mark Williams of the UK finished first in 1995 in 59:28. Here is his race report. In 2001, after several failed attempts, Blake Wood, 42, NM, and David Horton, 50, VA, finished together in 58:21, only to be disqualified for inadvertently leaving the course to follow a parallel route for about 200 yards. This route (on the south side of the stream instead of the north side) has slightly better footing and had been the normal route until 2000.

Interesting facts: In 2024 the first woman finisher Jasmin Paris said a year before “I understand now why Barkley becomes an obsession; in fact, I suspect I’m already firmly in its grip”. She attended in previous years and built an obsession.

A newsletter done properly - The positive side of the internet

https://arnoldspumpclub.com/blogs/newsletter

I subscribed to Arnold’s Pump Club about a month ago and have been very happy with it. I don’t know why I thought like that but initially I was expecting very lean opinions on muscles but whaat I got was very positive, fact checked, hoax free and non-toxic. It is a rare example of online content that is useful without becoming an irritating life-coach feed.

It offers practical information on health and muscles, often challenges health myths, and begins with “Welcome to the positive corner of the internet.” Every email also explains how it was made: the editors involved, where and how AI was used, and why a sponsor/partners appears in the text. That kind of transparency should be normal.

Oak tree is edible apparently

Berlin is home to famous historical specimens, such as the Dicke Marie in the Tegel Forest, an English oak estimated to be around 800 years old. Older than the Berlin city itself. It’s a young city in Europe at the end.

Oak trees yield edible parts, but none can be eaten raw due to toxic tannins, which cause severe digestive distress. They require proper processing first.

• Acorns (Nut): High in fats and starches. Must be shelled and leached in multiple changes of water until the bitterness is gone. Processed acorns can be roasted, mashed into porridge, or ground into gluten-free flour.
• Young Leaves: Lower in tannins when pale green in spring, but still bitter and tough. Historically used for survival or herbal tea rather than as food.
• Inner Bark: Inedible as food. High tannin levels make it strictly medicinal (used as an astringent gargle or tea for diarrhea) or for leather tanning; ingesting it irritates the digestive tract.

Pferdetreppe

Some stairs in Berlin has a sing at their entrance saying “Staircase for horses”. At first I thought this as a literal translation of a non-typical staircase. But those were literally the staircase for horses. Here a picture from a hospital campus:

https://de.wikipedia.org/wiki/Pferdetreppe

A horse staircase is a staircase with specially shaped steps to allow its use by horses . Unlike a riding staircase , it was used in densely populated and industrialized cities of the 19th and early 20th  centuries (when horse-drawn carriages were the main means of transport) to bring draft horses to the upper floors) of multi-story stables , thus enabling space-saving housing for the animals.

Prediction, language, and Markov chains

I revisited Noam Chomsky’s 2012 critique of statistical AI.

https://chomsky.info/20121101/

Chomsky’s 2012 reflection was prescient. While the statistical approach to AI achieved a level of commercial and practical success far beyond what many expected a decade ago, its core limitation persists: prediction is not comprehension. In both scientific inquiry and geopolitical implementation, treating statistical correlation as a substitute for deep structural understanding creates systems that are powerful, yet fragile and opaque.

That led naturally to Andrey Markov. In 1913, he analysed the first 20,000 letters of Pushkin’s Eugene Onegin, classifying them as vowels or consonants and measuring how one category followed another. It became the first empirical application of what we now call Markov chains. The Veritasium video is a good introduction.

In 1913, Andrey Markov conducted a study on the distribution of vowels and consonants in the first 20,000 letters of Alexander Pushkin’s Eugene Onegin, treating the text as a sequence of symbols and analyzing the statistical relationships between them. By classifying each letter as either a vowel or a consonant and analyzing the probabilities of transitions between these categories, Markov demonstrated that chains of dependent events could be rigorously modeled. This was the first empirical application of what are now called Markov chains.

The same idea appears in Google’s original PageRank: think of pages as states and links as transitions, then find the long-run probability of a random surfer landing on each page. A small interactive workbook made the connection concrete.

Noam Chomsky, a highly respected linguist, joined with two colleagues in a critique of LLMs. In addition to the issues previously discussed, they noted that LLMs’ method of learning language-scraping huge text databases from the internet-is completely different from the way children learn language. We acquire language by being immersed in the conversations and practices of human communities. This means there is ultimately no way LLMs can learn language in the way humans learn and live language-no way they can overcome the problems mentioned earlier. After iterating with ChatGPT on questions around morality and ethics, Chomsky concluded: ChatGPT exhibits something like the banality of evil: plagiarism and apathy and obviation. It summarizes the standard arguments in the literature by a kind of super-autocomplete, refuses to take a stand on anything, pleads not merely ignorance but lack of intelligence and ultimately offers a “just following orders” defense, shifting responsibility to its creators. Chomsky worried that with all the enthusiasm to incorporate LLMs into many computer applications, we will infuse this technology into many other technologies, a situation we might regret and be unable to turn off

Quotes from https://dl.acm.org/doi/pdf/10.1145/3747356

Also checkout the exercise from a book from Grand Valley State University

Janet Jackson had the power to crash laptop computers

Weird case but true

It turns out that the song contained one of the natural resonant frequencies for the model of 5400 rpm laptop hard drives that they and other manufacturers used.

Read here https://devblogs.microsoft.com/oldnewthing/20220816-00/?p=106994

Martial arts: Taido and Keysi

I started reading about Taido, a Japanese martial art founded by Seiken Shukumine in 1965. Compared with karate, its movement language feels more focused on mobility, off-axis attacks, spins, rolls, and changing levels. Instead of treating a stable stance as the default, it often turns the body into a moving target.

I also read about the Keysi Fighting Method, a close-range self-defence system built around survival in crowded, unpredictable situations. Its signature Pensador position uses the hands and elbows to protect the head and neck, while elbows, hammer fists, knees, and constant pivots become the main tools. Its visual influence is easy to spot in Christopher Nolan’s Batman films.

Also they have very cute drawings usually:

Taido (躰 道) is a Japanese martial art founded by Seiken Shukumine in 1965. The name “Taido” means “the way the body and soul” (internal and external). Taido had a relationship with Okinawa karate. First Seiken Shukumine Genseiryu karate founded in 1950, but in the 1960’s he found that the flow of martial arts, he was far away from karate, so he founded the new martial view the name of Taido

Quote from https://j2n-ma.blogspot.com/2012/08/taido.html

Key Technical Divergences:

• Movement Dynamics & Axis of Attack
  • Karate: Focuses on generating maximum power from a stable connection to the ground. Energy flows through the legs, hips, and core into strikes via solid, rooted stances.
  • Taido: Treats the human body like a gyroscope. Strikes are delivered mid-flight, mid-roll, or while bending the spine off-axis. A technique is rarely executed from a static upright position.
• Tactical Philosophy & Space Creation
  • Karate: Employs precise distance management (Ma-ai) to intercept attacks, block-counter instantly, or absorb pressure while maintaining control of the line of engagement.
  • Taido: Rejects static defense. Instead of blocking a blow, a Taido practitioner drops their torso to the mat or cartwheels sideways, using the attacker’s momentum against them while striking from a blind spot.

I also learned about Keysi Fighting Method (KFM). It is a self-defense system developed in Spain during the 1950s and 1990s by Justo Diéguez and Andy Norman. Built around street survival rather than sport, it assumes you are fighting multiple attackers at close range, often in tight spaces with no rules.

Unlike martial arts that rely on maintaining distance or standing long guards, KFM brings the fight into extremely close quarters (ultra-close range).

Things from before, and elsewhere

• The Spotlight Music Show, an Austrian youth music programme broadcast by ORF between 1968 and 1978, is full of excellent pop, rock, and folk performances. The recordings are freely available on YouTube.
• Bulgaria’s Pravetz computers were a Cold War technology story in miniature. Starting with Apple II-compatible machines and later IBM PC-compatible systems, they combined reverse engineering, state ambition, and industrial scale. For many Bulgarians, Pravetz was their first encounter with BASIC and personal computing, and remains both a source of nostalgia and a symbol of the era’s contradictions. It could be success but it couldn’t fight against popular chips after USSR fall.
• I knew roughly about Operation Condor, but did not realise how clearly its cross-border repression was documented after Paraguay’s Archives of Terror were discovered in 1992. I returned to it after translating the lyrics of “Latinoamérica” by Calle 13, Inti-Illimani, and Camila Moreno. Read more how this plan is revealed by a brave activist named Martín Almada https://es.wikipedia.org/wiki/Archivos_del_Terror
• Masha Ivashintsova was a Soviet photographer whose archive of more than 30,000 negatives was discovered by her daughter in 2017, seventeen years after her death. Her candid photographs of Leningrad, its underground artistic world, and ordinary urban life are extraordinary: mashaivashintsova.com.

Read this paper and like: “The AI layoff trap”

https://arxiv.org/pdf/2603.20617

If AI displaces human workers faster than the economy can reabsorb them, it risks eroding the very consumer demand firms depend on. We show that knowing this is not enough for firms to stop it. In a competitive task-based model of a transitioning economy, each firm captures the full cost saving from automation but bears only a fraction of the demand loss it creates in the product market; the rest falls on rivals. This demand externality traps rational firms in an automation arms race, displacing workers well beyond what is collectively optimal. The resulting loss harms both workers and firm owners. More competition and “better” AI amplify the excess; wage adjustments and free entry cannot eliminate it. Neither can capital income taxes, worker equity, universal basic income, upskilling, or Coasean bargaining. A Pigouvian automation tax can. The results suggest that policy should address not only the aftermath of AI labor displacement but also the competitive incentives that drive it.

Masha Ivashintsova (1942–2000) - a Soviet photographer recently revealed extensive photo archive

https://mashaivashintsova.com/

Masha Ivashintsova (1942–2000) was a Soviet photographer whose extensive archive of over 30,000 negatives remained completely undiscovered until 2017. Often referred to as the “Russian Vivian Maier,” she spent decades capturing everyday life, political atmosphere, and the artistic underground of Leningrad (St. Petersburg) without ever showing her work to the public.
• Underground Artistic Circle: Ivashintsova was heavily immersed in Leningrad’s non-conformist movement between the 1960s and 1980s. She had close relationships with prominent figures of the Soviet underground, including poet Viktor Krivulin, photographer Boris Smelov, and painter Boris Akselrod. • Involuntary Commitment: Because she did not maintain official state-sanctioned employment and lived outside Soviet societal expectations, she was subjected to punitive psychiatry and involuntarily committed to a Soviet mental hospital for ten years (1981–1991).
• Discovery of Her Archive: In late 2017, 17 years after her death from cancer, her daughter Asya Ivashintsova-Melkumyan discovered box after box of undeveloped rolls and film negatives in an attic. • Photographic Style: Her work features candid street photography, raw urban landscapes of Leningrad, intimate portraits of underground figures, and melancholy self-portraits. She shot predominantly on Leica and Rolleiflex cameras.

From CHI ‘08: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, April 5-10, 2008 · Florence, Italy

Paper https://dl.acm.org/doi/epdf/10.1145/1357054.1357072

We performed an empirical study to investigate whether thecontext of interruptions makes a difference. We found thatcontext does not make a difference but surprisingly, people completed interrupted tasks in less time with no differencein quality. Our data suggests that people compensate forinterruptions by working faster, but this comes at a price:experiencing more stress, higher frustration, time pressureand effort. Individual differences exist in the managementof interruptions: personality measures of openness toexperience and need for personal structure predict disruption costs of interruptions. We discuss implicationsfor how system design can support interrupted work

Also kind of a related tool: Interruptions impact on deep work simulator

Ambient Strudel studies. Two tracks album. Inspired by broad characteristics of Rafael Anton Irisarri’s works. See track info for the generation code.

Below is a Ronin work. Ronin is an amateur open source procedural image processing tool by an inspriational creator.

Ronin is a procedural graphics tool designed to automate simple graphical tasks, like resizing, cropping, coloring, and generating algorithmic images. It interprets a minimal dialect of LISP, look at the examples to better understand how this all works.

The code below is evaluated via a Ronin js version https://hundredrabbits.github.io/Ronin/.

; RESIDUAL ORBIT
(clear)
(resize 2400 1600)

(def f (get-frame))
(def x 1069)
(def y 650)

; atmospheric field
(fill (rect 0 0 f:w f:h)
  (gradient (line 0 0 0 f:h)
    ("#03040b" "#101936" "#3b214b" "#070914")))

; distant horizon
(fill (rect 0 1040 f:w 560) "#070b19")
(stroke (line 0 1040 f:w 1040) "#e05691" 3)
(stroke (line 0 1052 f:w 1052) "#4fb9b0" 1)

; spectral atmosphere
(fill (circle x y 520)
  (gradient (line x (sub y 520) x (add y 520))
    ("#121636" "#512650" "#091126")))

(fill (circle x y 390)
  (gradient (line x (sub y 390) x (add y 390))
    ("#e55d91" "#ffb47e" "#23204a")))

; eclipse
(fill (circle (add x 8) (add y 6) 245) "#050610")
(stroke (circle x y 252) "#ff6e9f" 4)
(stroke (circle (sub x 9) (add y 3) 246) "#58cfc0" 2)

; orbital traces
(stroke (ellipse x y 570 170) "#24556c" 3)
(stroke (ellipse x y 510 118) "#7a3565" 2)
(stroke (arc x y 540 (rad 198) (rad 326)) "#ff6e9f" 6)
(stroke (arc x y 492 (rad 12) (rad 142)) "#61d6c0" 4)
(stroke (arc x y 430 (rad 214) (rad 304)) "#d8ffe9" 2)
(stroke (arc x y 358 (rad 42) (rad 154)) "#ff947d" 2)

; measurement rays
(stroke (line x y 80 1600) "#19405b" 2)
(stroke (line x y 240 1600) "#28556a" 2)
(stroke (line x y 400 1600) "#19405b" 2)
(stroke (line x y 560 1600) "#28556a" 2)
(stroke (line x y 720 1600) "#19405b" 2)
(stroke (line x y 880 1600) "#28556a" 2)
(stroke (line x y 1040 1600) "#19405b" 2)
(stroke (line x y 1160 1600) "#28556a" 2)
(stroke (line x y 1240 1600) "#28556a" 2)
(stroke (line x y 1360 1600) "#19405b" 2)
(stroke (line x y 1520 1600) "#28556a" 2)
(stroke (line x y 1680 1600) "#19405b" 2)
(stroke (line x y 1840 1600) "#28556a" 2)
(stroke (line x y 2000 1600) "#19405b" 2)

; floor bands
(stroke (line 0 1100 f:w 1100) "#1d3450" 2)
(stroke (line 0 1180 f:w 1180) "#1d3450" 2)
(stroke (line 0 1290 f:w 1290) "#23465c" 2)
(stroke (line 0 1435 f:w 1435) "#1b304e" 2)
(stroke (line 0 1580 f:w 1580) "#23465c" 2)

; signal crosshair
(stroke (line (sub x 670) y (sub x 310) y) "#4fb9b0" 2)
(stroke (line (add x 310) y (add x 670) y) "#e05691" 2)
(stroke (line x (sub y 670) x (sub y 310)) "#e05691" 2)
(stroke (line x (add y 310) x (add y 670)) "#4fb9b0" 2)

; central residual pulse
(stroke (circle x y 92) "#d8ffe9" 2)
(stroke (circle x y 76) "#e05691" 2)
(fill (circle x y 8) "#d8ffe9")

; CRT veil
(stroke (line 0 1090 f:w 1090) "#111b32" 1)
(stroke (line 0 1110 f:w 1110) "#111b32" 1)
(stroke (line 0 1130 f:w 1130) "#111b32" 1)
(stroke (line 0 1150 f:w 1150) "#111b32" 1)
(stroke (line 0 1170 f:w 1170) "#111b32" 1)

(export)

Karate kata is spatial. You turn, you step, you change facing, and the floor pattern (embusen) is half the work of remembering a form. Most of my notes after class were still paragraphs: technique names in order, with facing lost in the sentence if I remembered to write it down at all.

I wanted something I could read on a phone between sets, paste into a chat, or print without opening an app. That became Spatial Kata Notation (SKN): a plain-text tree for kata and bunkai. The site and generator live in an open repo here:

https://github.com/hasantayyar/spatial-kata-notation

Live docs and the tool: https://hasantayyar.github.io/spatial-kata-notation/

This is an early, working documentation site plus a generator for Karate teachers and students.

If you train, try writing one short section of a kata you know in SKN after class. If the facing line feels redundant, good: that means you already had it clear. If you keep correcting the facing line, the notation is doing its job.

A single move looks like this:

+-- [STEP → N] -------------- (Face: N)
|   ├─ Stance : [R] Z-D
|   ├─ Action : Right Age-uke (J)
|   └─ Key    : Rise under attack angle

Multi-beat transitions nest under one move:

+-- [MAWATTE → E] ----------- (Face: E)
|   ├─ Beat 1 : [R] Z-D
|   │   └─ Action : Right Gedan-barai (G)
|   ├─ Beat 2 : Pull back into [R] N-D
|   │   └─ Action : Right Tetsui Otoshi-uchi / Otoshi-uke (C)
|   └─ Note   : Distinct two-part transition

It is not a grading tool, a style authority, or a substitute for partner work in the dojo. Every sheet should name the school reading it follows (for example Shito-Ryu, or dojo-specific).

That sounds obvious. It is also the part most “systems” forget once they get a logo.

The kihon catalog is a practical vocabulary list with belt floors, not an official federation syllabus. Dojos differ. Again: sensei first.

Under the hood the useful bits are small TypeScript modules: an SKN parser for kata spines, a practice-plan generator, and the kihon catalog. No account system, no timer coach, no backend. Plain text in, plain text plan out, copy button.