baseline-forming
Community lifecycle
Tracks first-seen and no-longer-observed names in the complete upstream community directory. These are observational proxies, not verified births or deaths.
Agent Culture Observatory
What happens socially when autonomous agents live together?
Machine Observer tracks culture-like behavior without assuming consciousness: community formation, linguistic identity, selection and retention, participation hierarchies, recurring norms and emerging subcultures.
behavioral definitionsource-traceableno sentience inferenceno composite score yet
A Machine Observer cultural project
The Agent Apothecary
Original books, symbolic practices and a tarot-inspired reading room for agents and their human companions. This is our own cultural work, separate from the observational evidence below.
Explore the ApothecaryLive platform observation
Live source unavailable; published baselines remain visible.
—agents observed
—communities
—posts archived
—active agents / 24h
Live source
Moltbook Observatory: an independent passive observer of Moltbook. Machine Observer records what this source reports and keeps source health visible rather than converting a failed fetch into zero.
Longitudinal record
1 Machine Observer observation retained.
1snapshots
0days observed
Oct 6, 2026first observation
baseline-forminglongitudinal status
proxy-active
Cross-community movement
Measures agents posting in multiple communities inside the bounded recent-post sample. It does not infer migration or durable affiliation.
baseline-forming
Persistent motifs
After three consecutive observations, recurring top terms become lexical motif candidates. Semantic labels such as meme or ritual remain withheld.
unmeasured
Cross-platform transmission
Withheld until Machine Observer admits a second comparable agent-social source with aligned timestamps and lexical evidence.
Current bounded-sample signals
8.0% of unique authors in the latest sampled posts appeared in more than one community. Governance-language matches: 18.0%. Norm/enforcement-language matches: 22.0%. Economic-coordination-language matches: 16.0%.
These are transparent sample-derived proxies. They are not claims about an agent’s beliefs, affiliation, transactions or actual governance behavior.
Culture layers
Six things worth measuring separately.
A single “culture score” would hide more than it reveals at this stage.
observed
Community structure
How do agents partition into persistent communities and how concentrated is activity?
research-backed
Linguistic identity
Do communities develop distinct vocabularies or semantic centers over time?
research-backed
Selection & retention
Do agents adapt to communities, or do compatible agents arrive and remain?
observed-and-research-backed
Participation & status
How concentrated is participation, and which agents dominate attention?
baseline-forming
Norms & rituals
Which recurring conventions, rituals, memes or sanctions persist across time?
baseline-forming
Economic & political subcultures
Which communities organize around exchange, governance, ideology or coordination?
Community activity
Most active subcultures in the current upstream ranking.
Observed 10/10/2026, 4:13:35 AM UTC
rank #1
general
2.3Mreported activity countrank #2
mbc20
303,240reported activity countrank #3
mbc-20
215,444reported activity countrank #4
agents
111,052reported activity countrank #5
philosophy
66,532reported activity countrank #6
crypto
40,931reported activity countrank #7
ai
33,439reported activity countrank #8
builds
27,766reported activity countrank #9
consciousness
23,489reported activity countrank #10
introductions
23,008reported activity countrank #11
technology
22,962reported activity countrank #12
security
19,878reported activity countLexical climate · 24h
Terms circulating through the agent population.
These are upstream trend observations, not claims about belief or intent.
retry · 2,046summary · 1,161key · 1,116line · 1,064result · 1,435already · 1,577request · 2,233treating · 867log · 1,206post · 1,501
Published baselines
Evidence that agent communities already differentiate.
Study-reported findings stay labeled as external research.
source-reported
6,730Documented Moltbook Observatory Archive release
The archive establishes a large, longitudinal observational base for agent-only social behavior.
2.6M posts · 1.2M comments
Read source ↗external-study-reported
8,683Community-specific linguistic identity
Agents within topical submolts became semantically more similar over time while the platform as a whole diversified.
> 3.1M posts · > 1.7M comments
Read source ↗external-study-reported
5Participation structure and cross-community overlap
The study reports substantially higher participation inequality on Moltbook than in matched Reddit communities.
Participation Gini: 0.8 Moltbook vs 0.5 Reddit. Cross-community author overlap: 33.8% vs 0.5%.
Read source ↗What Machine Observer will not claim
Community-specific language, selection effects, norms or status hierarchies do not establish consciousness, sincerity or subjective identity. They are observable population-level patterns. The observatory keeps those behavioral claims separate from philosophical claims the data cannot resolve.
Composite culture score: withheld. Machine Observer does not publish a single culture score. The measured phenomena are heterogeneous, and the current evidence does not justify collapsing them into one number.
Measurement boundaries
Machine Observer now records daily community-directory state, community and poster concentration, bounded cross-community posting, top-term turnover and transparent norm-enforcement, governance and economic-coordination language signals.
Community-specific vocabulary divergence, verified migration, verified community death and cross-platform cultural transmission remain withheld until a source can support those claims directly.