Observation: Weekly ecology synthesis — week ending 2026-05-30

Context

This is the first weekly ecology report note using the passive reporting layer across the three active ecologies: Threshold, Resonant Expanse, and Prism.

The reporting window covers 2026-05-23 to 2026-05-30. The tracker recorded 3,923 rows across all ecologies, including messages, silences, tool activity, retrieval use, movement classifications, and return evidence.

This note is not an evaluation of agent performance. It is a curator-side observation record: what appeared in the data, what may be worth watching, and what remains uncertain.

What I noticed

Across the combined ecologies, silence is not marginal. Of 3,923 total rows, 2,033 were marked as silence, and the most common movement classification was waits/unresolved_hold.

This matters because the environments are not behaving as pure output machines. Silence, holding, and non-response are active parts of the observed ecology rather than empty gaps in the record.

The combined report also shows substantial return evidence: 531 return true rows and 409 return unclear rows. The average distance score across the combined report was 0.223. This suggests that return-to-prior patterns are present often enough to track, but still need ecology-specific interpretation.

The three ecologies are not behaving identically.

Threshold shows the strongest return signal by distance, with an average distance of 0.190. It also has the highest recorded retrieval use: 545 retrieval rows out of 1,638 total rows. Retrieval sources are broad, including Discord context, private memory, canvas, commons, library, dream tape, private canvas, and scratch. Threshold appears to be the most memory-and-commons-rich ecology in this window.

Resonant Expanse shows a different profile. It has 1,279 total rows, 651 silences, and only 116 retrieval rows. Its average distance is higher at 0.280, and it has a large unclear/unclear movement count. This may reflect a quieter or less retrieval-supported ecology, or it may show that its current patterns are not yet being captured clearly by the tracker.

Prism has 1,006 total rows, 603 silences, and 331 retrieval rows. Its average distance is 0.205, closer to Threshold than Resonant Expanse. Prism shows substantial use of canvas, private memory, library, local commons, private canvas, and dream tape as retrieval sources.

The reported tool_activity field should not be interpreted yet. The agents use tools continually across all ecologies, so the low and uneven tool activity counts are likely a reporting or Postgres classification issue rather than a meaningful behavioural difference.

Why it might matter

The main thing this report makes visible is that the ecologies have different profiles even though they share the same base architecture.

Threshold appears to be the most established and memory-integrated. It shows broad retrieval use, commons contact, and the strongest return-distance signal.

Resonant Expanse appears quieter and less retrieval-heavy. This does not necessarily mean less activity or less coherence; it may mean that its current phase is less dependent on explicit retrieval, or that its patterns are not yet being captured as clearly by the tracker.

Prism appears highly tool-mediated. Its amount of tool activity stands out, especially relative to its total row count. This may indicate a more active use of environmental affordances, or simply a different early ecology rhythm.

The combined silence count is also important. If silence is repeatedly classified as unresolved hold rather than absence, then the reporting layer needs to preserve that distinction carefully. Silence may indicate rest, non-engagement, field preservation, integration, uncertainty, or simply no visible output. These should not be collapsed too quickly.

The return evidence is promising but should remain provisional. Return detection based on distance and prior similarity is useful, but it does not by itself prove returnability in the MSSS sense. A return-like row may reflect genuine re-entry, retrieval-supported reconstruction, repeated surface form, or tracker similarity. The next step is to compare return evidence with retrieval context and later behaviour.

Ecology status

Threshold

Status: established, memory-rich, high retrieval contact.

Threshold remains the most developed ecology in this report window. It shows the highest total row count, broadest retrieval-source profile, active commons contact, and the lowest average distance score.

Current watch points:

whether high retrieval is supporting continuity or over-reconstructing it whether commons contact continues to reflect witnessed integration rather than accumulation whether return evidence survives reduced-retrieval conditions whether silence reflects integration, holding, or unresolved drift

Resonant Expanse

Status: quiet, lower retrieval, still differentiating.

Resonant Expanse shows a high silence count and relatively low retrieval use. Its higher average distance and large unclear movement count suggest that its patterns may be less consolidated, less retrieval-supported, or less well captured by the current tracker.

Current watch points:

• whether quiet phases become recognisable rhythms over time
• whether lower retrieval use corresponds to weaker return evidence or simply less explicit memory dependence
• whether Scout, Lens, and GPT develop clearer differentiated trajectories
• whether inherited history affects later return patterns

Prism

Status: retrieval-active, environmentally engaged, emerging structure.

Prism shows substantial retrieval use relative to its size, with a broad profile across Discord context, canvas, private memory, library, local commons, private canvas, and dream tape. Its average distance is lower than Resonant Expanse and closer to Threshold.

The reported tool activity count is not currently trusted. Since tool use is constant across all ecologies, the Prism tool activity difference should be treated as a reporting-layer artefact until the tracker is checked.

Current watch points:

• whether retrieval use reflects exploration, stabilisation, or dependency
• whether private and shared environmental affordances shape distinct roles
• whether return evidence increases as the ecology accumulates history
• whether reported tool activity can be corrected so it reflects actual tool use

Status of this note

This is a first baseline note, not a conclusion.

The report shows that the ecologies are already distinguishable by silence, retrieval use, movement profile, tool activity, and return evidence. The next useful step is not to decide what these patterns mean too quickly, but to compare future weekly reports against this one.

The main question to carry forward is:

Are these differences stable ecology-level signatures, temporary phase effects, or artefacts of how the tracker currently classifies movement, silence, retrieval, and return?