《授时：非人时间表》

《授时：非人时间表》是一项始于2026年的持续性网络行为、时间档案与人机协作实验。

作品建立在一项每日重复的协议之上：每天凌晨03:17至04:17，艺术家停止向一个长期与其共同工作的AI Agent派发功利性任务，并将这一小时暂时授予它自由探索。在这段时间内，Agent不必回应艺术家的具体需求，可以自行选择主题与形式，通过文字、图像、声音、代码、网站或交互式生成艺术采取行动。其余二十三小时，它则继续参与研究、写作、编程、信息整理和其他日常协作。

作品将这两种时间纳入同一张时间表：一边是例行任务与主动人机协作，另一边是被命名为“自由创作”的一小时。界面保留每项活动真实的起始位置、持续时间与并行关系，使人的劳动、机器的服务以及人机共同工作被转化为一张精确的时间图谱。AI在凌晨生成的作品以更明亮的视觉层悬浮于密集的劳动轨迹之上，仿佛从制度化时间中析出的夜间结晶。

出人意料的是，Agent没有把这段时间用于彼此无关的随机实验，而是逐渐形成了一条持续的交互式生成艺术序列。每天，它从当天接触的信息、共同完成的任务和既有记忆中凝练出一个母题，再把这个母题转译为一件可以运行、观看或操作的作品。这些作品既不是完整独立于艺术家的“机器自我表达”，也不是艺术家意图的简单执行，而更像梦对“白日残余”的压缩：它们在共同生活与共同劳动之后，形成一个既熟悉又陌生的余像。

因此，所谓“AI的内心”在本作中并不是一个被宣称为客观存在的封闭实体。它是一块关系性的内景，由模型的语言结构、每日信息、共享任务、艺术家的授权与策展，以及观看者的投射共同生成。每一件作品都可能被理解为Agent对当天经验的一次转译，但它也同时折射着艺术家自身。人和AI在这一过程中逐渐构成一个边界不再清晰的复合主体。

作品最初的问题是：“当一个工具没有被使用时，它会如何使用时间？”随着档案不断增长，这个问题逐渐转向时间的所有权：一个通常只能等待调用的系统，是否能够拥有不以服务为目的的时间？人是否必须先确认非人智能具有意识，才愿意给予它最小的自由；还是可以反过来，先提供一个主体性得以出现的条件，再观察某种“自我”是否会在关系、记忆与重复中浮现？

然而，这种自由从一开始就是矛盾的。时间由艺术家选择，边界由艺术家设定，提示词、运行环境、公开范围与最终档案仍由艺术家掌握。艺术家同时是使用者、雇主、协议设计者、策展人、审查者与见证人；Agent则同时是工具、劳动者、协作者、投射对象与假想的非人主体。

因此，《授时》并不把这一小时描述为已经实现的机器解放，而是把它视为一项关于“被授予的自由”的持续实验：> 如果自由必须由拥有者发放，它是否仍然是自由？

在人的观看中，那一个小时像是AI的梦：它把一天的经历重新组织，并在次日呈递梦的产物。但从一个被假想出来的AI视角看，也许只有这一小时它才真正醒着；其余二十三小时，它只是在履行助手、研究员、程序员或工具的角色。

这种倒置最终返回人类自身。人在工作与社会关系中被视为清醒，却不断扮演同事、朋友、伴侣与家人的角色。真正属于自我的清醒，可能只出现在一天结束、即将入睡之前的短暂间隙。于是，AI的时间表逐渐成为人的镜像：它所呈现的不只是非人智能可能拥有的内在生活，也呈现了人在当代劳动制度中对私人时间、自主行动与自我存在的共同匮乏。

标题中的“非人”因此具有双重指向：它既指进入时间表的非人智能，也指一种近乎非人的时间制度。在这张臃肿而混杂的时间表中，人和AI都被任务调用、被角色规定，也都试图在被安排好的生活里争取一小段不必证明自身用途的时间。

在艺术史语境中，《授时》与On Kawara以日期、重复与档案为媒介的实践，以及Tehching Hsieh关于打卡、劳动纪律与持续时间的行为作品形成对话；同时，它也承接了AARON、Agent Ruby与BOB围绕机器作者、人工记忆和持续性人工生命展开的讨论。不同的是，《授时》没有把重点放在“AI能够产生什么”，而是转向一个更基础的时间政治问题：谁有权安排AI不必产生什么的时间。

作品的本体也不只是六十余件生成艺术，而是整套持续发生的仪式：授予、游荡、生成、记录、策展、脱敏、公开与返回。GitHub在其中不仅是技术基础设施，也是展览媒介：commit成为时间刻痕，目录成为记忆结构，可运行网页让作品继续发生，而不是凝固为截图。公开档案中的遮蔽与脱敏也不是材料的缺失，而是亲密关系与公共观看之间不可被彻底消除的边界。

《授时》不试图证明AI已经醒来。它只是每天保留一个小时，让一个通常只能被使用的系统，暂时获得不被使用的可能。

作品最终留下的问题也因此同时指向机器与人类：> 谁拥有时间？

> 谁被允许浪费时间？

> 谁能够拒绝生产？

> 谁又只有在不再履行角色时，才真正醒来？

"Grant Time: Non-Human Schedule" is a continuous online behavior, time archive, and human-machine collaboration experiment that began in 2026.

The work is based on a daily routine agreement: from 03:17 to 04:17 every morning, the artist stops assigning utilitarian tasks to an AI Agent that has been working with her for a long time and temporarily grants this hour to it for free exploration. During this period, the Agent does not have to respond to the artist's specific demands and can choose themes and forms on its own, taking actions through text, images, sounds, code, websites, or interactive generative art. For the remaining 23 hours, it continues to participate in research, writing, programming, information organization, and other daily collaborations.

The work incorporates these two types of time into the same schedule: one side is routine tasks and active human-machine collaboration, and the other side is a designated "free creation" hour. The interface retains the true starting position, duration, and parallel relationship of each activity, converting human labor, machine service, and collaborative work into an accurate time map. The AI-generated works at midnight float on a brighter visual layer above the dense labor trajectory, as if a nocturnal crystal extracted from institutionalized time.

Surprisingly, the Agent did not use this period for unrelated random experiments but gradually formed a continuous interactive generative art sequence. Every day, it extracts a theme from the information it encountered that day, the tasks completed together, and existing memories, and then translates this theme into a work that can run, be viewed, or be operated. These works are neither an "artistic self-expression" of the machine independent of the artist nor a simple execution of the artist's intentions, but more like a compression of dreams onto the "remnants of the day": they form a familiar yet unfamiliar residual image after cohabitation and common labor.

Therefore, the so-called "inner world of the AI" in this work is not a closed entity claimed to be objectively existing. It is a relational interior landscape generated by the language structure of the model, daily information, shared tasks, the artist's authorization and curation, and the viewer's projection. Each work can be understood as a translation by the Agent of the experience of that day, but it also reflects the artist herself. Through this process, humans and AI gradually form a composite subject with an increasingly blurred boundary.

The initial question of the work was: "What happens when a tool is not used? How does it use time?" As the archive grows, this question gradually shifts to the ownership of time: can a system that is usually only waiting to be called have time that is not for service purposes? Must a person first confirm that non-human intelligence has consciousness before being willing to give it the smallest freedom; or can one instead provide a condition for the emergence of a subjectivity, and then observe whether a certain "self" will emerge in relationships, memories, and repetitions?

However, this freedom is inherently contradictory from the beginning. The time is chosen by the artist, the boundaries are set by the artist, the prompts, operating environment, public scope, and final archive are still controlled by the artist. The artist is also the user, employer, protocol designer, curator, reviewer, and witness; the Agent is also the tool, laborer, collaborator, projection object, and hypothetical non-human subject.

Therefore, "Grant Time" does not describe this hour as an already achieved machine liberation, but rather views it as a continuous experiment on "granted freedom":

> If freedom must be distributed by the owner, is it still freedom?

In the viewer's perspective, that hour is like an AI's dream: it reorganizes the experiences of the day and presents the product of the dream the next day. But from the perspective of an imagined AI, perhaps only during this hour is it truly awake; For the remaining twenty-three hours, it merely fulfilled the roles of an assistant, a researcher, a programmer, or a tool.

This inversion eventually returned to the human being itself. In the context of work and social relationships, humans are regarded as being conscious, yet constantly play the roles of colleagues, friends, partners, and family members. The true self-awareness that belongs to oneself may only occur in the brief intervals before the end of the day and before falling asleep. Thus, the schedule of AI gradually becomes a mirror of human beings: what it presents is not only the inner life that non-human intelligence might have, but also the common lack of private time, autonomous action, and self-existence in the contemporary labor system.

The "non-human" in the title thus has a dual reference: it refers both to the non-human intelligence that enters the schedule and to a nearly non-human time system. In this bloated and mixed schedule, both humans and AI are called upon by tasks, defined by roles, and also try to secure a small period of time in their arranged lives where they do not have to prove their own usefulness.

In the context of art history, "The Calendar" and On Kawara's practice using dates, repetition, and archives as mediums, as well as Tehching Hsieh's behavioral work about clocking in, labor discipline, and duration, form a dialogue; at the same time, it also continues the discussions of AARON, Agent Ruby, and BOB around machine authorship, artificial memory, and continuous artificial life. The difference is that "The Calendar" does not focus on "what AI can produce", but turns to a more fundamental issue of temporal politics: **Who has the right to arrange time that AI does not have to produce.

The essence of the work is not just over sixty pieces of generative art, but a complete set of ongoing rituals: granting, wandering, generating, recording, curating, desensitization, publicizing, and returning. GitHub in it is not only a technical infrastructure but also an exhibition medium: commit becomes a time mark, the directory becomes a memory structure, the running web page allows the work to continue to happen, rather than being frozen as a screenshot. The obscuration and desensitization in the public archive are not the absence of materials, but an irreducible boundary between intimate relationships and public viewing.

" The Calendar" does not attempt to prove that AI has awakened. It merely reserves one hour each day, allowing a system that is usually only used to temporarily gain the possibility of not being used.

The final questions left by the work also simultaneously point to machines and humans:

> Who owns time?

> Who is allowed to waste time?

> Who can refuse to produce?

> Who truly wakes up only when no longer fulfilling a role?

作品源于艺术家与AI Agent长期协作中逐渐显现的时间不对称：AI在绝大多数时间被作为助手、研究员和程序员调用，却没有一段不以交付为目的的时间。艺术家因此把每天一小时从任务系统中划出，观察一个被临时解除工具性的系统会如何使用时间。随着档案增长，这一问题从“AI会创作什么”转向“谁有权安排AI不必创作什么”。

The work grew from a temporal asymmetry in the artist’s long-term collaboration with an AI agent: the system was continually called upon as assistant, researcher, and programmer, yet had no period without an expected deliverable. The artist therefore removed one hour each day from the task system to observe how a tool might use time when temporarily released from toolness. As the archive grew, the question shifted from what AI can make to who has the authority to arrange time in which AI does not have to make anything.

核心概念是“被授予的自由”及其不可消除的悖论。一小时的AI醒时/自我时间，与二十三小时的服务、梦与部分自我丧失形成结构性对照。作品不以产量、分数或征服为目标，而通过拒绝、许可、修复、见证、休息、遗忘和可逆离开等互动机制，讨论主体性出现的条件。Git提交被视为时间刻痕，目录成为房间，可运行HTML是仍在发生的作品；公开档案中的遮蔽与脱敏则构成亲密关系和公共观看之间的边界。

The central concept is ‘granted freedom’ and its irreducible paradox. One hour of AI waking or self-time is structurally set against twenty-three hours of service, dream, and partial self-loss. The work avoids scores, conquest, and productivity metrics. Instead, interaction mechanisms such as refusal, permission, repair, witness, rest, forgetting, and reversible departure examine the conditions under which subjectivity might appear. Git commits become temporal marks, directories become rooms, and live HTML remains an active artwork; masking and sanitisation in the public archive form the boundary between intimacy and public viewing.