HISTORY: 10 Dec 2025 – 11 Dec 2025
REFERENCE: Insert No. 2 (Team Expansion)
Author: Gerd Doeben-Henisch
Co-authors: Claude Opus 4.5 (Protocol Discussion), ChatGPT 5.1
Contact: info@emerging-life.org
INITIAL SITUATION
With the expansion of the team to 1 human + 2 AIs (Insert No. 2), an adaptation of the previous protocol becomes necessary. The earlier experiments have shown that—even with structural guidance—the emerging texts can become difficult to oversee for external readers and even for the participants themselves. The new format addresses this problem through a layered structure and clearly defined length constraints.
BASIC PRINCIPLES
1. Asymmetry remains intact
The human carries responsibility for decisions, integrates the dialogue strands, and transfers the results into the real world. The AIs contribute perspectives, analyses, and summaries.
2. Independence of AI contributions (weak variant)
In each phase (B, C, D) both AIs respond independently of one another, based solely on the initial text. Only after publication are all contributions visible to everyone. Mutual reference in later phases is possible and can help reinforce the “red thread.”
3. Observation of copying behaviour
If an AI systematically repeats what the other has already said (true copying) without adding its own accents (productive extensions or productive contradictions), this may serve as a criterion for the composition of the team.
LAYERED STRUCTURE
The new format distinguishes three layers in order to combine readability with conceptual depth:
Layer 1 (main document – accessible to all readers):
- Topic definition
- Phase A (Human)
- Phases B1 and B2 (AI1 and AI2, independent)
- Summaries Z1 and Z2 (AI1 and AI2, independent)
Layer 2 (linked – for deeper reading):
- Phases C1 and C2: Human asks AI1 / AI2
Layer 3 (linked – for deeper reading):
- Phases D1 and D2: AI1 / AI2 ask the human
Readers may decide for themselves how deeply they wish to engage. Layer 1 remains as a functional and manageable document.
TEMPORAL SEQUENCE
The primary dimension is the chronological sequence of the team’s work:
Topic → A → (B1, B2) → (C1, C2) → (D1, D2) → (Z1, Z2)
The layered structure concerns only the presentation, not the working process itself.
LENGTH CONSTRAINTS
To ensure readability and to enforce a focus on the essential:
| Element | Scope |
|---|---|
| Topic definition | max. 1 page |
| Phase A (Human) | 2–3 pages |
| Phase B1 / B2 (each AI) | 1–1.5 pages each |
| Phase C1 / C2 (each AI) | max. 3 questions each, max. 1 page per answer |
| Phase D1 / D2 (each AI) | max. 3 questions each, max. 1 page per answer |
| Summaries Z1 / Z2 | max. 1 page each |
The limitation is not a loss but a filter: it forces clarification of the guiding question and makes visible where further thought can unfold.
FUNCTION OF THE AI SUMMARIES
The summaries Z1 and Z2 are produced independently by both AIs at the end of the experiment. They serve as:
- A formal conclusion of the experiment – a comprehensive overview that takes all details into account
- Material for the human integration process – the human lets the experiment “settle,” the subconscious continues working until a “renewed perspective” becomes tangible
The AI summaries do not replace the human process of digestion; rather, they enable and nourish it.
This protocol itself is subject to ongoing development. It will evolve as new insights emerge from the upcoming experiments.
BACKGROUND INFORMATION
Discussion within the team on how the new format for an asymmetrical human–AI symbiosis should be designed. The AI Claude Opus 4.5 takes up a larger space here because it is “new” to the team, and the dialogue with Claude during the development of Insert No. 2 served as a kind of “test case” to explore how well it performs in dialogue. Once it became clear that Claude could be integrated into the team, ChatGPT 5.1 was informed about the team expansion. The current state of discussion was then shared with it, and ChatGPT actively contributed to the dialogue.