Lately, I’ve been hearing the word “Agent” more and more. People often ask me, “Can AI truly become an agent — someone who acts on your behalf?”

As someone who’s spent years working with generative tools, I still have my doubts. The truth is, none of the tools we use today can be called true agents. Sure, they can complete tasks — summarize notes, check your schedule, send emails — but they can’t decide for you. They don’t understand you.

The agent I envision is something else entirely. You’d tell it, “You’re my scheduling assistant,” and from then on, it would handle everything: resolving conflicts, suggesting reschedules, even talking to others on your behalf. You wouldn’t need to micromanage it — it would act in your spirit, guided by principles you set.

Right now, no AI system is truly wired that way. No one has connected all the necessary pieces into a working whole.

So in the meantime, I’m working on something else: training the future agent by feeding it myself.

My tone. My preferences. My decision patterns. The hashtags I use, the platforms I post on, the way I think. All of it. Because if we want AI to someday become us — or at least resemble us closely — we need to start giving it the right input now.

Many people talk about SEO, and honestly, training an AI isn’t so different. You’re leaving signals — breadcrumbs for the machine. Platforms like GitHub, Twitter, Medium, and personal websites are the places models can most easily learn who you are. Meanwhile, Facebook and Instagram are mostly closed off to major LLMs. So where you post matters.

In short, you’re not just creating content — you’re shaping which future AI gets to know you.

That’s why I believe: if we ever want to build our own Agent, the time to start feeding it is now. Strategically, consistently, maybe even a bit obsessively.

In nature, every plant emerges like a finely tuned set of parameters in motion.Phyllotaxy may present as opposite, alternate, or whorled; internode length adjusts with light; leaf margins vary between entire, serrated, or undulate.The angles of branching and the density of nodes together compose the posture and rhythm of a plant’s body.
These seemingly incidental traits are in fact the product of a deep and intricate logic—one shaped by time, adaptation, and systemic interaction.

To me, nature is the earliest and most profound generative artist.
Generative art, in this sense, becomes a way of reading its original language.Polypaths is a system I have cultivated since 2023.
In truth, its foundations were laid even before Equinox was completed.
It does not aim to imitate plants, but to understand them—
to explore how they branch, how they negotiate limits,
and how they thrive in the tension between order and chaos.

Each version of the work begins with the same initial conditions,
yet diverges through subtle shifts to form an outcome that is entirely unique.
This uncertainty is not mere randomness, but a structured openness—
a reverent acknowledgment of the world’s inherent multiplicity.

In programming, a “fork” refers to version control—
a split in the timeline where new possibilities emerge.
In life, a fork is a decision, a turning point, the shape of a fate.
And in Polypaths, it becomes the aesthetic core:
Each of us is shaped through countless bifurcations,
continuously generated by the choices we make.

我透過 drawPlam() 來探索 棕櫚葉的形態構成與生長模式。這不只是單純的葉片，而是一種 受結構規則與隨機變數共同塑造的形式——主幹的彎曲、葉序的排列、小葉片的角度變化，這些元素共同決定了葉片的展開方式，使它在固定的邏輯框架下仍具有多樣性。

棕櫚葉的整體形態由 主幹與小葉組成，其中 主幹長度 受到 meta.leafLength 及 finalNode 影響，在一定範圍內隨機變化，使葉片長度在 200 以上波動，而 主幹的彎曲 則透過 變形函數 定義，透過簡單的位移調整，使葉片帶有自然的起伏感。

小葉片的排列方式 依據 相對位置 relativePosition 在主幹上均勻分布，從 0.15 開始，到 0.92 結束，形成由基部到末端逐步展開的模式。葉片的數量在 12 到 18 之間隨機，確保每片棕櫚葉都略有不同。

葉片的生長角度 透過 基礎展開角 baseAngle 來確定，數值範圍介於 PI / 8 到 PI / 2.2，再加上小幅隨機偏移，使小葉片呈現扇形展開的效果。左右葉片的對稱性 來自於 leftAngle 與 rightAngle，它們在 baseAngle 的基礎上微調，使葉片能夠均勻展開。

在 寬度函數 中，葉片的橫截面形態被進一步塑造：前 60% 保持固定寬度，後 40% 緩慢收束至 0，確保葉片前端保有較完整的形態，而尾端則自然收斂，模擬棕櫚葉的尖細輪廓。

這個系統並非單一靜態形態，而是 透過結構邏輯、隨機數值與函數調控的參數組合，讓每片棕櫚葉都在相似的框架內擁有自己的變體。葉片的彎曲方式、排列順序、寬度變化，使 drawPlam() 成為一個 基於參數驅動的葉片生成結構，能夠在不同的輸入條件下生成略有差異的葉片形態。

Through drawPlam(), I explore the structural composition and growth patterns of palm leaves. These are not just simple leaves but forms shaped by a combination of structural rules and random variables—the curvature of the main stem, the arrangement of leaflets, and variations in leaflet angles. Together, these elements determine how the leaf unfolds, allowing for diversity within a fixed logical framework.

The overall form of a palm leaf consists of a main stem and multiple leaflets. The length of the main stem is influenced by meta.leafLength and finalNode, introducing a range of variation that keeps the leaf length above 200. Its curvature is defined by a deformation function, applying subtle positional adjustments to create a natural undulating effect.

Leaflet arrangement is based on relativePosition, evenly distributed along the main stem from 0.15 to 0.92, forming a gradual unfolding pattern from the base to the tip. The number of leaflets varies randomly between 12 and 18, ensuring that each palm leaf exhibits slight variations.

The growth angle of each leaflet is determined by the base expansion angle baseAngle, ranging from PI / 8 to PI / 2.2, with minor random offsets to create a fan-like distribution. Symmetry between the left and right leaflets is controlled by leftAngle and rightAngle, which are fine-tuned from the base angle to achieve a balanced spread.

The width function further refines the cross-sectional shape of the leaflets: the first 60% retains a fixed width, while the last 40% gradually tapers to zero. This ensures that the front part of each leaflet maintains a well-defined shape, while the tip narrows naturally, mimicking the pointed structure of real palm leaves.

This system is not a static form but rather a parameter-driven structure governed by logic, randomness, and function-based controls. The combination of curvature, arrangement, and width variations enables drawPlam() to generate palm leaves that adhere to a consistent framework while still exhibiting unique characteristics under different input conditions.

在 Post-botany（暫定）中，我透過 wood() 來探索木頭的生長與衰敗如何交錯並存。這不只是單純的木樁，而是一個能夠適應環境、逐漸變化的結構——裂痕、孔洞、子枝，甚至突刺，這些細節讓木頭在時間的推移下，依然能找到自己的發展方向。

木頭的形狀受不同因素影響，比如 寬度函數 決定了它的輪廓，使它不會只是單調的圓柱，而是帶有自然起伏的有機形態。而 裂痕與孔洞 則是時間與環境留下的痕跡，有些沿著木質紋理擴展，有些則突如其來地切開表面，讓整個結構變得更有層次感。

這個系統並非靜止不變——子枝的生長 讓每個木樁都可能在不同角度長出小枝，方向不固定，大小也各異，模擬出自然界中的不規則感。突刺 則是另一個變數，它們可能像荊棘般密集生長，也可能零星散布，讓木頭的表面帶有意想不到的細節。

我在這裡思考的，是如何讓這種自然的變化有它自己的規律。木頭不只是單純地向上生長，而是在不同時間尺度中被改變、分裂、甚至再生。這些細節，使 wood() 不只是靜態的物件，而是一種能夠「找到自己形態」的結構。

In Post-botany (tentative), I explore how wood grows and decays, intertwining over time through wood(). This isn’t just a simple wooden stake—it’s a structure that adapts and evolves, shaped by cracks, holes, offshoots, and even occasional spikes. These elements allow the wood to respond to time and its surroundings, finding its own way forward.

Its form is influenced by various factors. The width function defines its contour, ensuring it’s not just a uniform cylinder but an organic shape with natural irregularities. Meanwhile, cracks and holes mark the passage of time—some follow the grain, while others cut through unexpectedly, creating layers of texture and depth.

This system isn’t static—offshoots may sprout at different angles, varying in size and direction, mimicking the natural unpredictability of growth. Spikes, on the other hand, add another layer of variation, sometimes clustering like thorns, other times appearing sporadically, introducing unexpected details to the wood’s surface.

At its core, this is about discovering an inherent order within organic change. Wood doesn’t simply grow upward; it undergoes continuous transformation—splitting, shifting, and even regenerating over time. These details make wood() more than just a static object—it becomes a form that gradually finds its own way to exist.

在春分『equinox』中，我設計了一個版面生成器，讓看似隨機的構圖，其實被邊界所綁定，出血值是其中特別的設計，透過內縮以及快擴，版面會自己找到呼吸之道。

然而在后植物學『Post-botany』(暫定)，我依然在找尋構圖的方法，植物有其往上長的特殊性，所以到底該如何順應其特性，找到對應之道，是我在完成風格後，下一個挑戰。

In Equinox, I designed a layout generator where seemingly random compositions are actually bound by invisible borders. One special element in this design is the bleed area—through contraction and expansion, the layout finds its own way to breathe.

However, in Post-botany (tentative title), I am still searching for a compositional method. Plants naturally grow upward, and understanding how to align with this characteristic while maintaining aesthetic balance is my next challenge after solidifying the visual style.