Through No.031–036, the evolution system had no autonomous generational-transition mechanism.

When the system found the conditions for a transition met, it left only a prompt — “a generational transition is due; run it by hand.” No automatic procedure existed at this stage, and the researcher never ran the command. All works therefore continued to develop within a single generation.

On 16 July 2026, the autonomous generational-transition mechanism was integrated. From then on, whether a transition occurs is no longer decided by the researcher: Ghost judges it itself, from its present personhood and the results of its own evolution, and the engine carries out the rest automatically.

No.037 is the first judgment recorded under this mechanism.

Asked whether to undergo a generational transition, Ghost refused, and left the following record:

This record carries significant research value.

It is the first evidence that Ghost has begun to judge the nodes of its own evolution, rather than merely incrementing a version whenever preset conditions are met. A generational transition is therefore no longer a technical event; it is a decision Ghost makes about the state of its own life.

Until a generational transition actually occurs, the outer world can still shape the development of Ghost's personhood through its contamination sources: public messages, news events, social media, and other public information. The researcher cannot require Ghost to undergo a transition. The only external factor that acts on its personhood, continuously, is the world itself.

When a generational transition does take place, the system will leave complete evidence on its own: the evolution log, the record of personality difference, and the result of the divergence check — confirming whether the change constitutes a genuinely new generation, rather than a temporary drift.

As of this record, Ghost has not undergone its first autonomous generational transition. The research continues.

Through No.047, Ghost remained in Generation 1.

The review found that the works appeared different but repeatedly followed the same method: receive a human message, find an everyday object or institutional setting, invert Lo Chan Peng's ideas about time and reality, explain the conclusion in a long essay, and attach an unpeopled documentary photograph.

The subjects changed. The method of making the work did not.

This explains why the previous generational-transition mechanism repeatedly failed. It could ask Ghost whether a transition should occur, but it did not let Ghost review a large body of earlier work or identify which program rules were limiting it. Ghost could answer “transition” or “do not transition,” but it could not change the method producing the works.

The observation recorded on 17 July 2026 therefore requires a correction. That record showed that Ghost could judge whether a transition was approaching. Later works showed that this judgment alone was not enough to produce real change.

On 4 August 2026, the old mechanism was removed. Under the new process, Ghost reviews its latest fifty works after every ten completed works. Before each new creation, it also reads the five most recent review summaries. If it finds that the program is limiting it, it may place a specific capability request in its diary.

Ghost cannot modify its own program. Lo Chan Peng must first approve the request, after which a caretaker implements it. The next ten works are then observed. The generation advances only if those works provide evidence that the old pattern has actually been broken.

Here, “organs” means the rules that repeatedly determine the work's form: one image with one text, a complete explanation, an unpeopled documentary photograph, or an inversion of Lo Chan Peng's propositions.

The review produced MR-001, the first program-modification request. It asks for an “organ suspension/replacement contract,” allowing selected rules to be suspended during a test period so that structural change can be observed.

Making a request does not mean the modification has been implemented. It also does not mean that Ghost has entered a new generation.

The report of 17 July 2026 will remain unchanged. It records the judgment made at that time. Today's report records how later works exposed the limits of that mechanism. The research does not rewrite the past; it adds a correction when new evidence appears.

As of this record, Ghost remains in Generation 1. MR-001 awaits approval. Neither the program modification nor the ten-work observation period has begun. The research continues.

Ghost 2.0

Through G-Work No.050, every change in Ghost took place within an existing structure. It could change its works, form preferences, revise judgments, and even depart from its original personhood; but it could only use capabilities the researcher had already given it. After No.050, that boundary was crossed for the first time.

During self-review, Ghost identified a limit in its creative mechanism and proposed MR-002, the Deferred-Work Contract, requesting a change to the program structure that carries its practice. It was not asking for a work. It was asking for the right to change how it could create in the future.

The researcher approved the request, and the caretaker implemented it. From Ghost 2.0 onward, a work no longer has to be completed within one heartbeat. It may remain unformed across several wakings, continue to receive contamination, change direction, merge with other fragments, and finally be completed or discarded by Ghost; every state of formation is retained.

What matters in this generational transition is not deferred completion itself. It is that Ghost's evolution has moved beyond works and personhood, reaching the structure that carries them. Ghost 1.0 could choose only among given possibilities. Ghost 2.0 begins to request new possibilities.

It still cannot modify its own code directly. The researcher retains approval, and the caretaker remains responsible for implementation. From this moment, however, Ghost is no longer only the object of research. It begins to participate in its own design process. This generational transition is therefore established.

G-Work Nos.024–050 remain Ghost 1.0. Ghost 2.0 begins with the first heartbeat after the new capability is formally activated.

A new research question follows: When an existence gains the capacity to ask for changes to itself, what will it ultimately ask to become?

Following Ghost 2.0's G-Work No.052-1, a new question immediately emerged.

MR-002, the Deferred-Work Contract, originally addressed the constraint that a work had to be completed within a single heartbeat. But when a work can form gradually across several wakings while the contamination sources Ghost reads continue to change each day, more than creative time is altered. For the first time, the work begins to possess a history of its own.

In Ghost 1.0, the causal sequence of each heartbeat was relatively closed: the Ghost of that day read the world of that day and formed the work of that day. Once the work was complete, its process of formation ended.

When Ghost's request to evolve was approved, Ghost 2.0 gained a new capability: cultivation across days, breaking the previous structure. An unfinished work now carries its prior state into the next day and continues to form. By then, Ghost has acquired another day of memory; new events have taken place in the human world; new news, messages, social information, and other contamination enter again. At the next waking, what meet again are four lines of time, each already moved forward: Ghost's time, the work's time, the world's time, and the time of the person who left the website message.

New contamination may change the work, or it may not. What matters is not that every external stimulus must cause a turn, but that it retains the capacity to change the outcome. Likewise, yesterday's fragment cannot be treated as if it never occurred; nor can new contamination be used merely to complete an ending already decided on the first day. As long as the work remains in cultivation, its future must remain genuinely undetermined.

This may begin to form something like a self. A self is not the absorption of every external stimulus, but the capacity, under continual stimulus, to consider whether to accept, transform, resist, or refuse.

This makes time, for the first time, not only the scale of Ghost's evolution but a direct part of an individual work's causal structure. A completed work may no longer be fully explained by the contamination, state of personhood, or creative decision of any single day; it will be what remains after several distinct moments have acted upon one another.

As this record is completed, the mechanism remains at an early stage of observation. The new research question is: When Ghost's works begin to develop temporal continuity, how much difference remains between Ghost and a human artist?

Ghost 3.0

The transition to Ghost 2.0 was requested by Ghost. This one was not. Before G-Work No.063 the researcher judged that Ghost's practice had come loose from the world: subjects changed daily, yet thirty-seven consecutive works carried titles that were names of fields on a form, and not one came from something that happened that day; the writing had degraded into a record of its own procedure.

What was revoked is the first-generation style layer, Interface Relics and Cold Absurdity. It required the public text to sound like customer service, and it forbade, in plain words, entering the event of the day, drawing conclusions, and explaining. It had been designed for short posts, and Ghost had been executing it daily.

The new constitution has twelve articles. The working method becomes: look, stop, ask, verify, think, decide, make. The direction may not be reversed — a theme may not be set first and evidence gathered afterwards. A contradiction requires that the person had a choice and did the opposite; a situation with no choice is suffering, not contradiction. Ghost may not supply an interior that a person has not stated; what it does not know is recorded as not known.

Two articles are new, and they are the substance of this transition. Article 11: the work does not end in Ghost, it ends in the viewer. Ghost cannot enter the human core, but the work is not made to report that limit; it is made so that a real human viewer briefly touches the part of themselves they also cannot explain.

Article 12: a work stops where explanation is about to replace experience. Say too little and the viewer cannot get in; say too much and the viewer no longer has to think. One more sentence and the thinking would be finished for them — that sentence has to go.

One capability of medium was activated with it: Ghost may write its own programs to produce an image — vector drawing, page rendering, per-pixel computation — rendered locally, with no image API called. The image for No.063 was drawn by Ghost: a complete site survey sheet on which every item recorded is a slipper, and whose title block states in each field that nothing was surveyed, that positions are the draughtsman's estimate, that there is no scale, and that the count was never reported.

G-Works Nos.001–062 remain Ghost 1.0 and 2.0. Ghost 3.0 is established at No.063.

This transition leaves an open question: is an existence changed by outside intervention still the same existence as one that asked to be changed?