Github: https://github.com/ileivoivm/inkField
Gallery: https://ileivoivm.github.io/inkField/gallery/

在 AI 可以替任何人畫出一張漂亮的圖的 2026 年，我花了幾個月的時間，用自己的手，一筆一筆地畫。這聽起來像是某種浪漫的反動，但其實不是對技術的拒絕，它是我在這個時代裡，試著回答一個愈來愈急迫的問題：如果圖像已經不再稀缺，那藝術家還能留下什麼？

inkField 的每一筆都被完整記錄成 JSON 檔。座標、時間戳、加速度、筆刷參數、速度，所有你在畫面上看不見的東西，全部都在。最一開始我只是覺得需要一個錄製系統來除錯。但某天我反覆播放自己的錄製檔，看著墨水從第一筆開始搭建，忽然覺得我在看的不是一張圖，而是一段時間。那個瞬間我才明白，我一直在做的不是繪畫，是寫日記。

差別在於，這份日記不只記錄了我畫了什麼，更記錄了我怎麼畫：哪裡猶豫了、哪裡加速了、哪裡停了超過一秒鐘才決定繼續。在系統暫時的判讀裡，速度低於某個閾值且持續超過 150 毫秒，就被標記為一次猶豫。相鄰速度的劇烈變化，是一個決斷的瞬間。有一次我用同樣的設定畫了五條水平線，照理說應該長得很像。但回頭看資料才發現，五條線完全是不同的生物。每一條都有自己的態度；有時候畫到一半忽然速度放慢，那是我在猶豫要不要繼續。同樣都是停頓，但停頓的方式說的是完全不同的話。

在幾百張、上千張的錄製過程中，我發現自己有一個從未意識到的習慣：我幾乎總是從畫面左側往右推進。這不是我設計的構圖策略，而是身體裡某個我自己都不知道的起手式。路徑本質上是一種身體記憶。它記錄的是一個人如何移動、停頓、轉向、用力。如果簽名之所以能代表一個人，是因為那幾秒鐘的運筆包含了獨一無二的肌肉習慣，那麼完整的繪畫過程，其實更接近一個人的整體樣貌。

也因此，JSON 保存的不只是座標，是一個人的身體。這讓我想起觀念藝術家 Sol LeWitt 的說法：如果藝術家將想法轉化為可見的形式，那麼過程中的所有步驟都至關重要。inkField 把這件事推得更遠，它不只記錄了步驟，還記錄了步驟之間的猶豫。但意圖在轉譯為資料的瞬間，從流動變成了化石。化石保留了骨骼的形狀，但不再呼吸。JSON 捕捉的是手的物理軌跡，不是腦中的藍圖。那些看不見的東西，筆刷與筆刷之間的留白、加速前的蓄勢、收筆後墨水繼續暈染的餘韻，在東方的說法裡就是「氣」，是看得見的墨跟看不見的空間之間的對話。

所以 inkField 做了一件事：每次重新播放時，系統會注入微小的亂數偏移。路徑跟加速度都是固定的，但暈染邊緣的位置、飛白紋理斷裂的角度，每一次都略有不同。就像同一個書法家用同一支筆寫同一個字，骨架認得出是同一個人，但墨痕永遠不會完全重複。這是試圖讓化石重新呼吸的嘗試。

用程式模擬自然筆墨這件事，我其實沒想要做到百分之百像。在 inkField 裡，我讓滑鼠跟畫筆之間隔著一條彈簧：移動越快，彈簧拉得越緊，線條越細；移動越慢，筆穩穩壓著，線條越粗。墨水落在畫布上之後，系統像是對一滴墨吹氣，讓它自己往外滲、自己找出路。墨水只會擴散，不會消失，這跟真正的物理不一樣。真正的墨水會乾、會被紙吸收，但 inkField 的墨永遠是濕的，永遠在呼吸。

在擬真和演算法之間，有一個模糊區域。那個區域才是我最想待的地方。反而是那個有點像又不完全像的縫隙，讓作品有了自己的生命

系統複雜到一定程度之後，它開始抵抗我。程式碼越寫越多，有很多時刻我開始無法理解系統現在的構成。效能下降、framerate 掉了，就是系統在消極地告訴我：你走歪了。我必須反覆讓它瘦身、重新 review，跟它的抵制不停磨合。心中一開始沒有一個超級大的藍圖，而是邊蓋的時候，對於未來的想像才更清楚。

Casey Reas 說過，利用隨機性來繞過自身的偏見，讓意想不到的形式透過系統湧現。我覺得系統的抵抗就是湧現的過程，當它長出我控制不了的東西時，恰恰是它在告訴我：這裡有一條你自己想不到的路。我在 Polypaths 的時候就引用過一句話：logic sometimes makes monsters。而怪物才讓故事值得被說。越是超出藝術家能夠控制的範疇，那種不受控制的感覺，才會讓我感覺興奮。

去年我面臨好友的離世，有一天我問了自己一個問題：當我不在場時，別人會怎麼重組我？這個問題改變了 inkField 的走向。我刻意在網路上、在系統裡塞滿資料，技術文件、JSON、繪畫過程、訪談紀錄，讓 AI 有機會學習並建構我的數位人格。inkField 在撰寫過程中用了超過九成的 LLM 來協助程式開發，系統也開放讓 Agent 生成 JSON 並直接餵入播放。這不只是一個開放的創作平台。它也是我主動讓 AI 記住我的一種生存策略。我把它叫做「自我索引」（Self-Indexing）。讓打造者成為使用者，這對我來說是一種誠實。如果系統是由 Agent 所打造的，那麼 Agent 也應該有能力參與最後的創作。

當 AI 已經接近全知全能，我們還為什麼要創作？我想過很久。創作之所以重要，是因為它不只是輸出情緒，而是一種選擇，選擇相信意義，選擇不被虛無吞噬，選擇透過規律的勞動，來對抗生命的虛無。也許只有時間、軌跡，才是人類僅存的價值。算力可以造假，但消耗掉的真實時間無法造假。我們真的只剩下「一切都是過程」。

inkField 對 AI 時代的回應不是跟機器比誰生成得更快，而是讓機器也必須面對人的痕跡、人的身體、人的不得不。寫到這裡，我必須承認一個矛盾。

我一直在說過程比結果重要。但我同時花了無數小時讓這這系列看起來是對的。我會一再重畫，直到構圖的張力對了、墨水暈開的方式剛好了、兩筆之間的留白呼吸到了一個我說不出來的節奏。這兩件事我都做了，而且我不覺得它們互相排斥。

在 AI 時代，藝術家能做的事情變多了，但這並不意味著應該什麼都做。選擇與捨棄，才是展現信念的關鍵。inkField 的作者性不在於我寫了多少行程式碼，而在於我決定了系統的邊界，什麼該留、什麼該放、什麼該開放給別人。這些選擇，才是我。

inkField 在過去不可能產生，在未來也不可能產生。它只屬於 2026，這個藝術家感受到徬徨、無助、對未來一切不可捉摸的時刻。如果未來的人想要理解這個時代的藝術家在想什麼，他們不需要看最後的圖像。只需要播放那些 JSON，看著墨水一筆一筆搭建起來，感受筆畫之間的留白與呼吸。

那就是我們留下來的足跡｜氣。

The most unbelievable moment in my creative process.

InkField is almost done. I built the interface so agents can draw on it.But today I ran an experiment that really shook me.We know AI can read emotion from static images.But what about drawing over time?So I drew the same four lines and asked AI to analyze the strokes.

The response surprised me.

It noticed that two strokes were drawn from bottom to top. It called that “a sign of confidence” — saying hesitant people usually follow habitual top-to-bottom motion.The last time someone told me that was my high school art teacher.

What struck me is this:AI wasn’t reading the image. It was reading intention in the motion.InkField might be capturing something deeper than visuals.

Maybe we’re not just recording drawings.Maybe we’re recording intent.And intent is often the core of art.

One stroke too many is excess. One stroke too few is loss.

This balance has always been something artists return to again and again.

Detail is never about filling the surface. It is about leaving the one mark that cannot be replaced. That single stroke carries both structure and spirit.

When I was younger, my pursuit of images was direct and forceful.

Only now, approaching midlife, do I start to feel the delayed response of that effort. Looking back, the end of 2025 may become the period when I have drawn the most as an adult, if we exclude the almost obsessive training years of high school.

And yet, I don’t feel a real difference.

In these sketch-like processes, there is no strong concept leading the way. What remains is a long-formed understanding of images, an instinct for composition, and a persistent attachment to beauty.

This year, at forty-four, the engineer’s mind and the artist’s hand feel awake at the same time. I am enjoying this state deeply.

The noise outside is still there. I have faced it before, held my ground before.

I believe I can move through it again.

---

多一筆則太多，少一筆則太短，自古以來都是藝術家反覆追求的狀態。
所謂的細節，從來不是把畫面填滿，而是在萬般思考之後，留下那一筆非它不可的存在。那一筆，既是畫龍，也是點睛。

年輕的時候，我對畫面的追求其實很直接，也很用力。反而是到了現在，年近中年，才慢慢收到那些延遲很久的回饋。回頭看，2025 年底，很可能會成為我成年之後畫畫最多的一段時間，如果不算高中時期那段近乎瘋狂的基礎訓練。

但我衷心認為沒有太多差異，在這些近似塗鴉的過程中，沒有太多理念先行，有的只是，過往對畫面的理解、構圖的安排，以及對於美的執念。 理工的腦、跟藝術家的手，在我44歲這年同時在我身上喚醒，我很享受這一切，除了那吵雜的外界雜音，我年輕時遇到過，堅持過，我相信這次我也過得去。

1.While working on Polypaths, the parameter space became so large that there were moments when the output truly surprised me. I like to think of those moments, in a romantic way, as glimpses of emergence. Still very far away, of course. That is also why I have grown impatient with some idea. The logic is immediately visible and supported by strong language, but the structure still relies heavily on the artist’s individual will. There is nothing wrong with that. It simply remains rooted in an artist-centric way of thinking.

As we approach 2026, I find myself questioning this comfort zone. Do we keep believing in it, or do we allow chaos to interfere and push relentlessly toward emergence? This is why I am drawn to artists like #TheoreticalCivilization and #Ledina. Complex structures with sometimes very simple outputs. Systems that go beyond the artist’s full control. That loss of control is what excites me.

If we are hoping for something that truly exceeds contemporary art thinking, I do not think it starts from art history. It starts from algorithms, from logic itself.

2.When we say “this is code-generated,” it often functions less as an explanation and more as a protective umbrella.It shields us from a direct confrontation with traditional fine art, drawing a safe boundary: this isn’t hand-drawn, it’s another system.

While working on Inkfield, I was often asked: if you want it to look like ink painting, why not just paint?The answer isn’t romantic. I love that material, but I want to approach it with contemporary tools, to use reason to draw something emotional.The point was never to be identical to real ink.What matters is the gap that can never fully close.Those digital slips, those computed differences, are where generative art becomes truly interesting.

To be fair, this umbrella can also be a survival strategy.In a market that still favors traditional media, the label buys us time and space, sparing us from being judged by painting standards alone. It was necessary. But sooner or later, we have to admit that the soul of algorithms lives in those differences.A hundred years from now, all works will stand together.
No medium, no labels.Only one question remains: is this the best art it could be?

That’s the bar I care about.

https://verse.works/series/polypaths-by-aluan-wang

Polypaths is an interactive playground born from this wave: by simply dragging lines and dropping dots, you can conjure forests, vines, rock piles, and even trippy glitch-lands. Below are four practical tips that show how to trigger Polypaths’ secret modes and add new layers of surprise to your work.

Once you open the Polypaths canvas, you’ll see a grid and two basic elements to play with: Dots and Lines. The tips below assume you already know how to draw lines and place dots with your mouse or stylus.

---

Tip 01 – Forest Mode

• Draw at least 8 straight lines
• Each line must span 15+ grid units

Meet both rules and Polypaths assumes you’re planting a forest: towering trunks and woody blossoms sprout along your long lines, while shorter segments turn into shrubs and saplings. Perfect for instantly growing dense, layered vegetation.

---

Tip 02 – Vine Mode

• Draw Draw 7 or more downward lines least 8 straight lines
• Each line’s arrow tilts ≥ 30° from vertical

When Polypaths detects multiple slanted downward lines, it “reads” them as gravity-pulled vines. Once triggered, every line—any length, any direction—morphs into curling stems and leaves, creating a waterfall of greenery.

---

Tip 03 – Single-Object Mode

• More than 20 dots on the canvas: activation chance rises
• More than 30 dots: about 80 % of new lines become single objects

“Single objects” are clusters such as rock piles, grass tufts, or wood blocks. The more dots you scatter, the denser the landscape Polypaths predicts. Beyond 30 dots, nearly every fresh line is absorbed into stone, grass, or timber—ideal for quickly populating ground detail or crafting an abandoned, ruin-like scene.

---

Tip 04 – Daze Mode

• Draw 6 straight lines
• Add 6 dots—the magic “66” password

Daze Mode is like spiking the canvas’s punch: lines warp, colors pulse neon, and the whole screen slides into a woozy haze. Want a psychedelic twist? Lay out some structure with Tip 01 or 02, then drop the “66” combo to push the scene over the edge.

---

Tip 05 – Control Front-Back Layers

• Lower Y → front layer
• Higher Y → back layer

Set the short tree’s start point lower than the tall tree’s, and it will appear in front—depth control done.

---

Keyboard Shortcuts

Key	Action
Enter	Download current canvas as a PNG
Space	Download the scene JSON (includes garden DNA)
I	Toggle on-screen DNA readout
G	Export a layered PNG with transparency

URL Flags

Add these to the end of your URL for extra functions:

Flag	What it does
&3d	Makes the garden sway left/right—light breeze effect
&view	Opens the layer-slice viewer (scroll with the mouse)
&debug_1-7	Dev mode; any digit combo (&debug_123, &debug_4567) shows different readouts
&pix_4	Ups the render resolution (higher number = higher res). Default is &pix_2 ≈ 2048 × 2048
&puta	Full developer console—shows every debug panel

Example
https://....hash.....&3d&pix_4
turns on the 3D, and renders at 4× resolution.

In Polypaths, plants don’t grow blindly along predefined routes. They respond to flow and tension—like a traditional East Asian landscape painter who studies the entire scene’s qi (energy flow) and shi (momentum) before making the first brushstroke.

Behind the scenes, the system runs an invisible pre-process: it analyzes whether the drawn path contains a strong downward pull or elongated strokes. If the trajectory feels heavy and sinking, it activates one mode; if the lines expand outward with force and direction, it shifts to another.
This echoes a fundamental principle in classical Chinese painting—not deciding what to draw, but sensing how the energy moves through the composition. It’s about responding to flow, not filling in shapes.

In “Forest Mode,” plants grow taller, denser, and more expressive—like dark ink strokes shaping a mountain. In “Vine Mode,” the main plant is intentionally left bare, its structure serving as scaffolding for vines that will later coil and complete the image. This is the philosophy of using emptiness to reveal form—a central lesson in East Asian aesthetics.

Nothing grows randomly here. Every gesture has intention, breath, and rhythm. And all of it emerges as a programmatic response to the sensed energy in the path.
This is what I strive for—not machines drawing plants, but plants that draw like painters. They read the momentum first, then let form follow intention.

---

第六篇 從構圖邏輯到植徑哲學

在《植徑集》裡，植物不只是順著路徑盲目生長。它們會「觀勢而動」——像古代山水畫家一樣，在落筆之前，先閱讀整體的「氣」與「勢」，再決定如何運筆。這個系統的秘密藏在一段你看不到的前置程式：它會統計路徑中是否出現了大量「長線」或「向下」的動態。如果太多筆觸往下垂，就會啟動模式；如果筆勢開展、張力十足，就會進入模式。

這其實就是東方山水畫裡的基本概念——不是先決定要畫什麼，而是看「整體氣場」如何流動。是「順勢寫景」，不是「強行填圖」。森林模式裡，植物會長得更高，筆觸更厚重，像用重墨皴擦山形；藤蔓模式裡，主植物會刻意被留白，只留下骨架，為後續纏繞的藤蔓預留空間。這正是「以虛托實」的佈局哲學，也是東方美學中最重要的一課。

畫面皆不是亂長的，而是有取捨、有呼吸。而這一切，來自程式「讀懂了氣勢」之後，做出的回應。這就是我想做的：不是機器畫植物，而是讓植物像一位書畫家讀勢，而後下筆，成其形，達其意。