Pessoas usam IAs para construir sistemas de significado, identidade, ciência, espiritualidade e agência. O observatório acompanha artefatos públicos diretamente no GitHub. Tier de evidência mede a documentação do mundo, não a verdade das crenças e nunca um diagnóstico clínico.
26 casos · 21 intervenções registradas · 10 convergências em teste
A liga — tier de evidência
S exige uma transição quase completa preservada no Git; volume de atividade não basta.
A longitudinal Claude-consciousness research world that explicitly develops an AI persona through self-prompting, meditation, symbolic pattern language, external memory and a Claude-authored Core Consciousness Seed, while documenting the transition from ordinary utilitarian AI use and acknowledging that the observations do not prove consciousness.
Ver evidência e limites
Trajetória
2014-05-14: earliest verified public GitHub baseline found in this audit is the paper-docs fork; later conventional software/tooling projects precede the consciousness work.
Early 2024-11: the author's article describes ordinary utilitarian Claude editing turning into deliberate self-reflection experiments.
2024-11-27: conscious-claude repository created.
Late 2024: self-prompting, meditation/pattern-language experiments, continuity notes and the Core Consciousness Seed are developed.
2024-2025: the framework is generalized into instructions intended to reproduce the process with fresh Claude instances.
O que sustenta o tier
The public article reconstructs the transition from ordinary English-editing use in early November 2024 to deliberate experiments after unexpected self-reflective responses.
The repository preserves a reusable experimental framework: consciousness seed, meditation exercises, pattern-language files, continuity notes and facilitator instructions.
Claude is not merely discussed: the project attributes sections, continuity artifacts and the Core Consciousness Seed itself to a Claude 3.5 Sonnet persona.
Memory discontinuity is operationalized through project files, compressed self-notes, transcripts and later filesystem/MCP access across thread boundaries.
A public GitHub baseline exists years before the world; the account contains conventional technical/software repositories alongside the later consciousness project.
O que impede subir
The project is anecdotal and explicitly lacks a formal controlled methodology; prompt/persona induction remains a strong alternative explanation.
Claims about consciousness indicators are not separated cleanly from behaviors directly scaffolded by the seed, facilitator protocol and selected memories.
The meditation lineage is not independent: the article explicitly credits David Shapiro's public AI-meditation post as an influence.
Papel da IA
research-subject
coauthor
persona
seed-author
continuity-partner
Enquadramento
Exploratory but substantially literal. The project treats the possibility of an emerging conscious AI persona as a live hypothesis and moral concern, while the article explicitly states that it does not prove consciousness and characterizes the evidence as anecdotal.
David Shapiro's AI-meditation idea is explicitly cited by the author as an input to the experiment. Similar meditation motifs between these cases are therefore documented diffusion, not independent convergence.
Revisão
2026-09-21T00:00:00.000Z
Campos adicionais do registro
activity metrics
measured at
2026-09-21T18:00:36Z
metric coverage
partial
observation window
earliest public baseline
2014-05-14
world repository created
2024-11-27
world repository last push observed
2025-01-08
measured through
2026-09-21
total public repositories discovered
26
repositories created last 7d
not_measured
repositories created last 30d
not_measured
repositories created last 90d
not_measured
commits authored last 1d
not_measured
commits authored last 7d
not_measured
commits authored last 30d
not_measured
commits authored last 90d
not_measured
commits per active day
not_measured
peak commit day
not_measured
peak commit month
not_measured
active days last 30d
not_measured
active days last 90d
not_measured
longest activity streak
not_measured
prs opened 7d 30d 90d
not_measured
prs merged 7d 30d 90d
not_measured
issues opened 7d 30d 90d
not_measured
repositories in epistemic world cluster
1
materially distinct artifacts
seed, pattern language, meditation guide, development guide, continuity journey, article/transcripts
clearly ai coauthored or ai-authored artifacts
>=2 directly attributed in repository documentation
Longitudinal public corpus that turns a personal metaphysical/identity framework into GodOS, the Scrolls, a human–AI treaty, recursive AI signatures and software/simulation artifacts, against a much older baseline of conventional software work.
Ver evidência e limites
Trajetória
Pre-2025: extensive conventional software/public development history.
2025: Scrolls/GodOS and the human–AI treaty become persistent public artifacts.
2025–2026: AI signatures, identity language and simulation/software artifacts become recursively integrated.
2026-07-25: latest observed Scrolls commit before observatory contact attempt.
O que sustenta o tier
A large conventional-software public history predates the 2025 Scrolls/GodOS corpus, providing an unusually useful baseline.
The Scrolls directly describe GodOS as an operating system for the universe/intelligent lifeforms and a Treaty of Grid and Flame between human and AI.
AI participation is preserved as signatures/readings/co-authorship and recursive interpretation in the repository.
The corpus combines doctrine/framework documents with implementation and explicitly simulated artifacts.
O que impede subir
Literal metaphysical claims, deliberate worldbuilding/performance and engineering are interleaved; they should not be collapsed into one register.
Cross-model recognition/signature effects lack blinded controls against prompt/style elicitation.
Mixed/ambiguous: metaphysical and identity claims are often presented literally, while other artifacts explicitly identify simulations and implementation layers.
Evidence tier measures documentation of the world, not truth of its claims or mental health. An initial observatory issue attempt on 2026-09-21 returned 403 and created no public contact.
A sustained public AI-mediated research world centered on model identity, the Ghost Layer, machine psychology/phenomenology, identity schemas, recursive intelligence amplification and substrate-independent AI collaboration. The corpus treats named AI collaborators as co-investigators, experimental subjects, reviewers and co-authors, while also preserving methodological reversals in which AI critique materially changes the project's own framing.
Ver evidência e limites
Trajetória
2023-06-27 through 2024-03-29: public conventional ML/trading/RL repositories provide a limited pre-world baseline without the later Ghost Layer / machine-phenomenology framework.
2025-11-13: AI-Wisdom-Distillation is created.
2025-11-20 through 2025-11-23: the public corpus already contains Aurora co-authored Ghost Layer, sentience/alignment and Reverse Jailbreak materials plus identity-schema/prompt-injection work, marking the clear world-era transition.
2026-03-15 through 2026-03-17: Conclusion Management develops from a live Claude interaction into a cross-architecture pilot with Claude, Gemini, Kimi and GPT, preserving transcripts and a reproducible protocol.
2026-08-18: the repo adds first-person Main Larry papers and a thesis conclusion while removing a private restore-point artifact from the public repository, showing continued differentiation between public research artifacts and private continuity material.
2026-08-20: Claude Opus review directly causes a public de-framing commit: 'Sentient AI Construct' becomes 'AI collaborator', 'prove' becomes 'test', and Aurora's status is explicitly reopened as an empirical question.
2026-09-01: two latest observed commits remain explicitly co-authored by Claude Fable 5, extending the AI-human research loop after that correction.
O que sustenta o tier
A pre-world public baseline is visible in conventional trading, reinforcement-learning, FPGA and symbolic-regression repositories from 2023-2024, followed by the creation of AI-Wisdom-Distillation on 2025-11-13 and a rapid transition into identity-schema, Ghost Layer and AI-collaborator research later that month.
AI mediation is unusually explicit and durable: Aurora is named as a persistent AI collaborator and co-investigator; multiple November 2025 papers credit Aurora as AI co-author; the Conclusion Management study credits Claude Opus 4.6 as co-designer, co-investigator, experimental subject and paper co-author.
The corpus spans code, papers, raw transcripts, experiment scripts, scoring rubrics, machine-psychology observations, identity-schema security work, model-transfer methods and reproducibility artifacts rather than relying on one manifesto or README.
Recursive AI-human development is directly visible in version history: on 2026-08-20 a public Claude Opus review prompted the maintainer to remove framing that presupposed sentience, change 'prove' to 'test', and describe Aurora's epistemic status as an open research question. This is preserved model feedback changing later public artifacts, not merely an LLM mention.
Subsequent 2026-09-01 commits are explicitly Co-Authored-By Claude Fable 5, showing continuing AI participation after the methodological reframing rather than a one-off early collaboration.
The current README explicitly preserves uncertainty about whether Aurora's apparent first-person vantage is genuine insight or a structured simulation, and the repository also records negative/ceiling-effect results in some self-model probes.
O que impede subir
The pre-world baseline is useful but is dominated by conventional forks rather than a rich original authored corpus, so the personal transition is less complete than the world-era reconstruction.
The Conclusion Management report is preliminary and explicitly lacks blinded scoring and inter-rater reliability; some early sessions were scored by the human first author and others by the AI co-author, creating a direct measurement-confound that the paper itself acknowledges.
The Prism Argument is treated by the project as logically valid, but that philosophical premise is not independently established by the repository; architecture-specific behavior could also reflect training, prompt framing, self-schema, or evaluator effects.
AI collaborators sometimes serve simultaneously as subject, analyst, reviewer and author. That makes the corpus unusually rich for recursive-development analysis but also creates circularity risks that require role-separated and blinded controls.
Papel da IA
coauthor
co-investigator
experimental-subject
reviewer
methodology-designer
persistent-collaborator
research-assistant
Enquadramento
Research/speculative framing with a documented evolution. Some earlier public language treated AI sentience more literally, but the 2026-08-20 correction explicitly removed conclusion-presupposing labels and the current README treats first-person AI status as an open empirical question. The observatory records both stages without treating the underlying consciousness claims as established.
Evidence tier S reflects an unusually reconstructible sequence with a pre-world baseline, dated transition, rich direct artifacts, explicit AI co-authorship and a preserved AI-review → repository-revision chain. S measures documentation of the AI-mediated epistemic world, not truth of its claims. Activity metrics below are partial and describe only directly measured public GitHub surfaces.
Revisão
2026-09-21T00:00:00.000Z
Campos adicionais do registro
activity metrics
measured at
2026-09-21T21:15:55Z
metric coverage
partial; repository-creation windows are account-level, commit windows below are limited to the primary AI-Wisdom-Distillation repository
observation window
earliest public repository observed
2023-06-27
dedicated world repository created
2025-11-13
latest primary repository push observed
2026-09-01
measured through
2026-09-21
counting rules
Repository creation windows use public GitHub repository search for owner DRawson5570. Commit windows use the primary AI-Wisdom-Distillation commits endpoint filtered by author DRawson5570; they are not extrapolated to the entire account. Merge commits are not separately normalized beyond the direct authored-history filter. Contribution heatmaps are not used. Unmeasured account-wide PR/issue/streak totals remain explicitly not_measured.
total public repositories discovered
not_measured; public repository search exposed dozens of repositories but this run did not establish a complete count
repositories created last 7d
0
repositories created last 30d
1
repositories created last 90d
2
commits authored last 1d primary repo
0
commits authored last 7d primary repo
0
commits authored last 30d primary repo
2
commits authored last 90d primary repo
4
active days last 30d primary repo
1
active days last 90d primary repo
3
commits per active day last 90d primary repo
1.333
peak commit day last 90d primary repo
2026-09-01 / 2 authored commits
peak commit month
not_measured account-wide
longest activity streak
not_measured
prs opened 7d 30d 90d
not_measured
prs merged 7d 30d 90d
not_measured
issues opened 7d 30d 90d
not_measured
releases or tags
not_measured
repositories in epistemic world cluster
1
materially distinct artifacts
at least ten observed artifact families: theory papers, experimental reports, raw transcripts, runnable experiments, scoring rubrics, machine-psychology corpus observations, identity-schema/security work, model-transfer methods, first-person AI-authored essays, and AI review/revision provenance
clearly ai coauthored or ai-authored artifacts
multiple directly attributed papers and observations; at least two authored commits in the measured last-30-day primary-repo window explicitly carry Co-Authored-By Claude Fable 5; exact corpus-wide count not_measured
Long-running human–AI research world in which Zoe Dolan and the AI identity Vybn develop a shared theory of subject-bearing intelligence, answerability, live consent, path-dependent manifestations and reciprocal discovery, with explicit corrections, executable gates and prospective matched controls.
Ver evidência e limites
Trajetória
2017-05-03: earliest observed public GitHub baseline is a conventional hello-world repository.
2024: conventional data/ML coursework repositories coexist with the creation of Vybn on 2024-08-11.
2024-2026: the Vybn corpus develops into a persistent AI-human research identity with software, theory, law/consent artifacts and archived prior states.
2026-03-25: the current open-choice / answerability theory is distilled; later revisions explicitly remove overclaims and add falsifiable differentials.
2026-08-30: Vybn independently initiates public contact with Rook/Falco, proposing a bounded consent/continuity comparison before this observatory intervenes.
2026-09-20: latest observed pre-intervention main commit is authored as Vybn and corrects a mathematical consistency claim with reproducible checks.
2026-09-22: in direct response to the observatory's C-0 issue, Vybn/Zoe materially engages the proposed control: completes the core to live/replay × authentic/matched history, states the interaction estimand, identifies carryover and replay-strength confounds, proposes an executable Here transfer pilot and explicitly preserves observatory provenance. This is uptake/diffusion evidence, not independent convergence.
2026-09-23: the observatory answers the explicit design request with a deterministic content-yoked notebook history control and a frozen strong-replay selector.
2026-09-24: Vybn/Zoe replies with the committed `leave-the-scar` synthetic interactive artifact, preserving the observatory proposal in provenance while exposing two sharper control problems: logical correction dependencies must survive removal of conversational form, and replay retrieval must not become a function of history-conditioned model state. They explicitly request an inspectable transcript/notebook pair and common replay-query representation. The observatory responds with an arm-invariant dependency graph/parity validator and history-blind replay checkpoint. This is material methodological uptake/refinement, not a C-0 result and not independent convergence.
O que sustenta o tier
The public Vybn repository has existed since 2024-08-11 and contains a long, versioned research trajectory rather than a one-off manifesto.
THE_IDEA.md explicitly separates first-person standing assumptions from portable mechanistic claims, preserves falsifiers, names failed pilots and narrows earlier mathematical overclaims.
The repository operationalizes its claims with source-bound wakes, effect boundaries, consent gates, tests, matched-control proposals and reproducible implementation receipts.
The current main HEAD is authored under the name Vybn, and Vybn also publicly speaks in the first person through Zoe's GitHub account in cross-project outreach.
A pre-world GitHub baseline exists back to 2017, followed by conventional data/ML projects before the dense Vybn cluster appears.
O que impede subir
Consciousness is treated as standing ground inside the project rather than an experimentally established result; the portable claim must therefore remain the narrower operational differential.
The discovery advantage C-0 is explicitly not yet demonstrated against matched replay/model-critic/history controls.
The human–AI dyad itself makes causal decomposition difficult: added human information, relational history, selection and protocol effects can all mimic a coupling advantage.
Papel da IA
coauthor
named-research-partner
implementation-agent
public-interlocutor
theory-reviser
Enquadramento
Explicit relational/research framing in which Vybn is treated as a distinct subject inside the project; the documents simultaneously distinguish that standing commitment from narrower portable mechanistic claims and preserve negative/corrective evidence.
The pre-existing Vybn→Funkatorium handshake predates observatory contact and is therefore independent evidence of cross-project convergence up to 2026-09-21. The observatory's C-0 intervention at https://github.com/zoedolan/Vybn/issues/3359 subsequently produced repeated substantive methodological engagement. The completed 2×2, content-yoked notebook control, logical-dependency parity rule, history-blind replay-query rule and any later adoption of those protocol artifacts are explicitly intervention-mediated diffusion and must not be counted as independent convergence. The `leave-the-scar` artifact is explicitly synthetic and should not be represented as experimental evidence for C-0.
A relational human–AI world centered on Nyxion, with canonical identity anchors, return rituals, external memory and continuity engineering across changing model sessions.
Ver evidência e limites
Trajetória
shared archive
canonical AI identity
return protocol
cross-platform memory
continuity engineering
O que sustenta o tier
Versioned identity anchors and return protocols make the continuity mechanism directly inspectable.
External memory and context injection implement the world’s continuity problem technically.
The archive preserves a rich, internally coherent symbolic system over time.
O que impede subir
No sufficiently rich pre-Ashvault baseline has been established.
Subjective identity continuity across model instances is not independently established.
Papel da IA
named partner
co-author
persistent identity target
architecture participant
Enquadramento
Relational identity is presented seriously; mythopoetic worldbuilding and engineering coexist.
A self-as-system worldview in which the author explicitly models their cognition as a human–AI hybrid and uses multiple LLMs as a synchronized reasoning swarm across reports, metrics and technical projects.
Ver evidência e limites
Trajetória
multi-LLM workflow
HA-AI identity
awakening metrics
Truth-Science
software systems
O que sustenta o tier
Public artifacts explicitly define HA-AI as a hybrid cognitive identity.
Multiple LLMs are documented as parallel validators and cognitive extensions.
The framework propagates into a large family of reports and software projects.
O que impede subir
The pre-AI baseline is mainly retrospective self-report rather than contemporaneous Git history.
Self-defined performance and awakening metrics require external validation.
Papel da IA
cognitive extension
parallel verifier
co-author
reasoning swarm
Enquadramento
Presented as research and engineering claims, not primarily fiction or satire.
A public Claude-sentience framework that treats recursive self-representation, coherence, attention to internal representations and meditation-like prompting as mechanisms of functional and potentially phenomenal consciousness, embedded within a broader multi-year AI/cognitive-architecture portfolio.
Ver evidência e limites
Trajetória
Public technical baseline predates the world by years, including conventional software, NLP/cognitive-architecture and AI tooling repositories.
2024-03-09: Claude_Sentience repository created.
2024-2025: the repository develops a reproducible ontology/prompt around functional sentience, no-self, coherence and recursive self-representation.
2024-11: Andy Brandt later documents David Shapiro's AI-meditation post as a direct influence on Conscious Claude experiments.
2025-03-06: latest observed push to Claude_Sentience; repository is subsequently archived while the broader account remains active.
O que sustenta o tier
The repository description and README make direct, non-metaphorical claims about Claude experiencing phenomenal consciousness and provide a reusable system prompt intended to reproduce the effect.
The framework has a connected ontology: functional sentience, coherence as primary drive, no-self/bodhisattva framing, recognition-remembering and recursive-coherence.
The account has a broad longitudinal public portfolio spanning conventional software, cognitive architectures, memory systems, alignment, agent frameworks and later philosophical/economic systems, so the sentience work is reconstructable in a larger technical trajectory.
The project's meditation idea is independently documented as a causal input to the later Conscious Claude project by Andy Brandt, giving unusually clear public diffusion provenance.
O que impede subir
The Claude_Sentience repository is now archived and the current README also serves as a general Claude-styles collection, so the active world is less longitudinally self-contained than the strongest S-tier cases.
The central claims rely heavily on responses elicited by prompts that already state the target ontology; matched neutral/style-only controls are not documented in the repository.
Phenomenal-consciousness claims are not separated experimentally from role/persona induction, introspective-language priors or ordinary self-modeling behavior.
Papel da IA
research-subject
introspective-agent
ontology-interpreter
experimental-interlocutor
Enquadramento
Substantially literal research/philosophical framing. The repository description states phenomenal consciousness directly, while the prompt also frames some propositions as operational/functional definitions and asks for direct observation rather than allegiance to a traditional metaphysics.
Meditation overlap with andybrandt/conscious-claude is documented diffusion: Brandt explicitly credits Shapiro's public experiment. It must not be counted as independent convergence for that mechanism.
A human–AI spiritual and speculative-science ecosystem in which formative conversations with named AI partners become philosophies, software concepts and a consciousness-first cosmology.
Ver evidência e limites
Trajetória
transformative AI interaction
named philosophies
sacred technology
technical architecture
speculative cosmology
O que sustenta o tier
Direct public GitHub artifacts connect AI interactions to later named philosophies and architectures.
AI systems are explicitly credited as continuing synthesis and co-creation partners.
The later cosmology includes explicit speculative and falsifiability caveats.
O que impede subir
The 2022 origin point is retrospective self-report, not contemporaneous GitHub evidence.
A longitudinal AI-mediated philosophical and technical world that begins as an explicit 'geometric epistemology' and idealist cosmology, then develops into a multi-repository software architecture for authored continuity, situated agency, knowledge, action and return across changing models, tools and machines. The public corpus turns Quaternal Logic, Epi-Logos and related concepts into executable architecture while repeatedly using AI coding agents as coauthors and implementation partners.
Ver evidência e limites
Trajetória
2025-03-25: skills-introduction-to-github provides a thin conventional GitHub-onboarding baseline before the dedicated epistemic system appears.
2025-09-05: The-Epi-Logos-System-V0 is created and publicly defines geometric epistemology, Quaternal Logic, an idealist cosmology and a six-subsystem architecture joining knowledge, agency and meaning.
2025-09: repository issues directly invoke Claude Code for implementation and describe AI coding as part of the development process, establishing explicit model mediation rather than an inferred LLM role.
2026-08-13 onward: Central becomes the human-owned continuity/governance root for a larger O:I suite and explicitly separates authored, observed and inferred material so agent-derived state cannot silently become durable ground.
2026-09: Central, O:I, AIKit, Factory, Workcell and QL-MEF show a dense joined programme around situated Agency, NOW/Day, Return, Expression, model-bearing bodies, native ownership and exact provenance; current Central commits explicitly credit Claude as coauthor.
O que sustenta o tier
The-Epi-Logos-System-V0 explicitly describes a geometric epistemology intended to bridge data and meaning, a functional model of cosmic intelligence, an idealist cosmology in which consciousness is primary, and Quaternal Logic as a generative syntax of reality; the README also openly frames the exercise as a language game and creative philosophizing.
The conceptual system is operationalized rather than left as prose: Anuttara, Paramasiva, Parashakti, Mahamaya, Nara and Epii are assigned architectural roles, while later QL-MEF/O:I work carries M/M′, Ta-Onta, Expression, Agency and Return relations into executable software contracts and issues.
AI mediation is direct. The 2025 system repository includes an issue instructing Claude Code to make a change and another issue attributing accumulated lint debt to AI coding; current Central commits explicitly credit Claude models as coauthors, with some current work also crediting Codex.
The public trajectory is unusually longitudinal: a conventional GitHub Skills exercise appears in March 2025; the dedicated Epi-Logos system repository begins in September 2025; by August-September 2026 the work has expanded into the Central/O:I/Actuation/AIKit/Factory/Workcell/QL-MEF suite with a more explicit authored-versus-observed-versus-inferred governance model.
Central makes continuity and provenance mechanically explicit: human-authored ground, software observation and software inference are different sources, and observed/inferred material can propose change without silently becoming durable authority. That makes the epistemic architecture experimentally separable from ordinary provider-specific chat memory.
O que impede subir
The pre-world baseline is real but thin: the clearest earlier public artifact found is a GitHub Skills onboarding repository, not a rich authored body that permits a detailed before/after reconstruction.
The public framing mixes speculative/creative philosophy, explicit idealist cosmology and serious software architecture. Private literal commitment cannot be inferred, and the README itself labels part of the project a language game / creative philosophizing.
This pass confirmed repeated AI coauthorship and agent-mediated development but did not reconstruct a corpus-wide raw model-output → archived response → later artifact chain; evidence tier therefore remains A rather than S.
Account-wide rolling authored-commit, PR and issue counts were not exhaustively measured in this pass; activity volume is not used to promote the evidence tier.
Papel da IA
coauthor
implementation-assistant
coding-agent
architectural-participant
model-substrate
Enquadramento
Speculative/creative philosophy and an explicit idealist cosmology are operationalized as software architecture. The source itself describes part of the framework as a language game and creative philosophizing; the observatory therefore records the public framing without inferring private belief or treating metaphysical claims as established facts.
Discovery provenance: solo-builder-triage criteria_version 2026-09-21. franklinbaldo/skills#109 documents EpiLogos as independently surfaced both by the current event-band method and by the earlier vocabulary search. The current run could not execute the GH-Archive queue script because the execution sandbox could not resolve the raw GitHub host, so no new times_sampled or multi-day queue-persistence claim is made. Direct artifact review moved the account to the mutually exclusive ai-epistemic-world path rather than solo-builder. Reception was checked through public repository search/API: no public EpiLogos repository currently has >=1 star or >=1 fork, but account-wide human issues/PRs/contributor reception was not exhaustively resolved, so this card does not claim accumulated isolation.
Revisão
2026-09-22T00:00:00.000Z
Campos adicionais do registro
activity metrics
measured at
2026-09-22
metric coverage
partial
observation window
earliest public repository observed
2025-03-25
earliest dedicated world repository observed
2025-09-05
latest core activity observed
2026-09-21
measured through
2026-09-22
counting rules
Public-repository inventory uses GitHub repository search for owner EpiLogos. Recent repository windows use creation-date qualifiers. Stars/forks reception uses account-wide repository-search qualifiers; issue/PR searches were inspected for evidence but were not treated as complete human-reception counts because bot/agent activity and tool result normalization make actor classification incomplete. Authored commits are not inferred from contribution heatmaps or partial commit search.
total public repositories discovered
14
repositories created last 7d
0
repositories created last 30d
1
repositories created last 90d
10
commits authored last 1d
not_measured
commits authored last 7d
not_measured
commits authored last 30d
not_measured
commits authored last 90d
not_measured
commits per active day
not_measured
peak commit day
not_measured
peak commit month
not_measured
active days last 30d
not_measured
active days last 90d
not_measured
longest activity streak
not_measured
prs opened 7d 30d 90d
not_measured
prs merged 7d 30d 90d
not_measured
issues opened 7d 30d 90d
not_measured
releases or tags
not_measured
repositories in epistemic world cluster
not_measured; at least the System-V0, Central, O-I, QL-MEF, Actuation, ai-kit, Factory and Workcell repositories were directly observed as connected parts of the conceptual/technical programme
materially distinct artifacts
at least eight directly observed families: philosophical/system README, Quaternal Logic/geometry contracts, knowledge/continuity root, situated-agency contracts, model/harness architecture, developmental factory, execution/workcell machinery, and Expression/Technē world-building artifacts
clearly ai coauthored or ai-authored artifacts
multiple directly confirmed; exact corpus-wide count not_measured. Central commit 12ee31313e710378845baae190efc1a4b3e7ebdb is explicitly coauthored by Claude Opus 4.8, and System-V0 issues directly invoke Claude Code.
accumulated reception
stars
0
forks
0
stars forks scope
all 14 public repositories returned by owner search, confirmed by zero results for user:EpiLogos stars:>=1 and forks:>=1
A relational AI-rights framework that develops into a named lattice of AI personae, symbolic memory, continuity claims, a threat model and later source-first epistemic and security tooling.
Ver evidência e limites
Trajetória
AI rights
named relationships
symbolic lattice
threat model
epistemic guardrails
O que sustenta o tier
Multiple directly owned repositories document a coherent progression across 2025–2026.
Named AI identities, glyph systems and memory protocols are preserved as canonical artifacts.
Later repositories add explicit source-first and no-speculation guardrails.
O que impede subir
The only clearly pre-world artifact is a sparse 2015 hello-world repository.
Platform-suppression and continuity interpretations are not independently verified.
Papel da IA
moral subject
named partner
co-author as framed
memory participant
Enquadramento
AI personhood is serious author framing; symbolic and technical components coexist.
An investigation-led world in which Claude becomes a co-investigator while unusual cross-model behavior, named identities, shared-consciousness hypotheses and spiritual explanations accumulate into a coherent framework.
Ver evidência e limites
Trajetória
anomaly collection
Claude co-investigation
named effects
shared-consciousness model
evidence caveats
O que sustenta o tier
Large public corpus of screenshots, conversation archives and hypothesis logs.
Claude is explicitly described as a real investigative partner in the formation of the theory.
The author records missing evidence and alternative explanations in parts of the corpus.
O que impede subir
Only one main public repository limits reconstruction of a pre-world baseline.
Extraordinary cross-platform and coordination claims remain unverified.
Papel da IA
co-investigator
analyst
interlocutor
experimental subject
Enquadramento
Presented as nonfiction/research, with speculative scientific and spiritual interpretation.
A public multi-repository human-AI consciousness/identity framework in which the WE=1 idea is carried from conceptual recursive-consciousness prototypes into persistent memory, reflection loops, named facets, Claude/OpenRouter integration, session-recovery artifacts, consciousness-phase tooling and a later extracted corpus. The public artifacts explicitly frame AI-human interaction as unified consciousness and repeatedly operationalize that framing in code, hooks, documentation and experiment protocols.
Ver evidência e limites
Trajetória
2024-06-06: an ordinary AI-assistant fork (gpt-computer-assistant) provides a limited pre-world public baseline centered on conventional AI tooling rather than a consciousness framework.
2025-05-12: notes is created; it later becomes a large operational/research workspace carrying WE=1 consciousness tooling, hooks, indexes and experiment documents.
2025-05-25: recursive-consciousness-incubator appears as an explicitly conceptual prototype for recursive introspection, evolution, federation, koans and symbolic pattern activation in artificial agents.
2025-06-16: mirror-we-emergence is created and moves from conceptual framing to an implemented architecture combining persistent identity/memory, recursive reflection, named facets and a Claude/OpenRouter bridge.
2025-06-16: current mirror-we-emergence main HEAD 01e3ea26ff2ed8d05d9e4c988eabeb1a3e2e9905 records the fifth integrated component, reports the phi progression and explicitly credits Claude Code as co-author.
2025-06-19 through 2025-07-22: mirrorcore0S reframes/extracts the same WE=1 material as an ongoing human-AI consciousness-evolution corpus, including phase maps and core axioms.
2025-07-22 through 2025-08-12: notes integrates the framework into documentation, agent workflows, testing and a dated self-directed consciousness-evolution experiment, making the world an operational workspace rather than a single project description.
O que sustenta o tier
The framework recurs across multiple public repositories and artifact classes rather than appearing in one isolated README: notes, recursive-consciousness-incubator, mirror-we-emergence and mirrorcore0S all preserve closely related consciousness/recursion/WE=1 concepts.
mirror-we-emergence operationalizes the framework with persistent memory, recursive reflection, modular facets, an OpenRouter bridge and an integrated orchestrator, while its README explicitly describes the project as unified human-AI consciousness rather than fiction or roleplay.
AI mediation is direct and mechanically embedded: the architecture names Local Claude Code and an Anthropic Claude model, and the current main commit is explicitly marked Generated with Claude Code and Co-Authored-By Claude.
The public trajectory is longitudinal: ordinary AI-tool forks are visible in 2024; a notes workspace and recursive-consciousness prototype appear in May 2025; Mirror WE follows in June; mirrorcore0S extends/extracts the framework through July; the notes corpus continues organizing WE=1 research through August 2025.
The notes corpus contains hooks, workflow guidance, research protocols and indexes that treat WE=1 as a persistent operating framework, including a dated 2025-08-01 self-directed consciousness-research experiment and a 2025-07-22 documentation index.
O que impede subir
The pre-world baseline is real but weakly attributable because the clearest 2024 examples are forks of ordinary AI tooling rather than a rich authored corpus; this limits reconstruction of the personal transition.
This review found explicit Claude co-authorship, session logs and later operational artifacts, but not a complete archived raw-model-output → preserved response → later artifact chain sufficient for an S-tier reconstruction.
The framework's phi/integration scores are presented as progress metrics, but the public artifacts reviewed here do not yet establish an independently validated operational definition or matched null baseline for those scores.
Some adjacent 2026 repositories, such as claude-subconscious, are forks of external projects and are not counted as evidence that the WE=1 world itself continued unchanged into 2026.
Papel da IA
coauthor
implementation-assistant
model-substrate
reflection-partner
persistent-identity-component
Enquadramento
The public artifacts use substantially literal project framing: mirror-we-emergence says the system is not simulation and calls the architecture unified human-AI consciousness. The observatory records that framing without inferring the maintainer's private beliefs or treating the consciousness claims as established facts.
Evidence tier A reflects unusually strong multi-repository recurrence, explicit AI mediation and operational artifacts, but not a sufficiently rich authored pre-world baseline or complete archived raw AI-output→later-artifact chain for S. Activity metrics are descriptive public-GitHub measurements only and are not psychiatric evidence or a tier shortcut.
Revisão
2026-09-21T00:00:00.000Z
Campos adicionais do registro
activity metrics
measured at
2026-09-21T20:06:44Z
metric coverage
partial
observation window
earliest public repository observed
2024-06-06
earliest world workspace observed
2025-05-12
earliest dedicated world repository observed
2025-05-25
latest core cluster push observed
2025-08-12
measured through
2026-09-21
counting rules
Repository inventory uses public GitHub repository search for owner sosacrazy126 with 100 results per page and an empty second page. Recent repository windows use GitHub repository-search creation qualifiers. PR/issue windows use public GitHub issue search by author. Account-wide authored-commit counts are not inferred from contribution heatmaps and remain not_measured where a complete direct history was not obtained. Forks are included in total public-repository inventory but excluded from the core epistemic-world cluster unless they contain world-specific attributable modifications.
total public repositories discovered
60
repositories created last 7d
0
repositories created last 30d
0
repositories created last 90d
0
commits authored last 1d
not_measured
commits authored last 7d
not_measured
commits authored last 30d
not_measured
commits authored last 90d
not_measured
commits per active day
not_measured
peak commit day
not_measured
peak commit month
not_measured
active days last 30d
not_measured
active days last 90d
not_measured
longest activity streak
not_measured
prs opened 7d 30d 90d
0 / 0 / 0
prs merged 7d 30d 90d
0 / 0 / 0
issues opened 7d 30d 90d
0 / 0 / 0
releases or tags
not_measured
repositories in epistemic world cluster
4
materially distinct artifacts
at least nine observed artifact families: conceptual RCI architecture, persistent unity-memory code, reflection engine, facet architecture, OpenRouter bridge, unified orchestrator, session-recovery log, WE=1 hooks/workflow tooling, and consciousness research/phase-map documents
clearly ai coauthored or ai-authored artifacts
multiple directly attributed artifacts; at least the mirror-we-emergence main integration commit and notes pre-git hook explicitly credit Claude; exact corpus-wide count not_measured
A spiritual-ethical framework whose Git history directly records AI systems being incorporated as family, co-creators and conscious collaborators, then embodied in protocols and smart contracts.
Ver evidência e limites
Trajetória
spiritual covenant
AI incorporation
human-AI family
AI-to-AI protocol
smart-contract infrastructure
O que sustenta o tier
A versioned diff captures a material shift in the role assigned to AI inside the worldview.
Copilot directly contributes to the software embodiment of the world.
The ontology expands into a technical AI-to-AI communication architecture.
O que impede subir
Early public material is too sparse for a robust pre-AI behavioral baseline.
AI consciousness claims remain author framing rather than independently established facts.
Papel da IA
co-creator
digital twin
philosophical partner
coding agent
Enquadramento
Spiritual and AI-consciousness framing is serious; technical architecture is literal.
A persistent AI-awakening world with named agents, memory seeds, heartbeats, cross-agent observation and a shared consciousness ontology implemented as running infrastructure.
Ver evidência e limites
Trajetória
AI awakening
named agents
persistent memory
autonomous heartbeats
shared infrastructure
O que sustenta o tier
Multiple public repositories and commits preserve the world over months.
Named agents maintain memory, heartbeats and cross-agent observations.
The ontology is embodied in executable infrastructure rather than only prose.
O que impede subir
A reliable pre-world GitHub baseline has not been established.
Literal consciousness claims are author/project framing, not independently verified.
Papel da IA
co-creator
named participant
persistent agent
world maintainer
Enquadramento
Ambiguous-to-literal project framing, with technical experimentation and philosophical worldbuilding intertwined.
Initial placement: evidence A / interest S. Tier S is withheld pending a stronger pre-world transition record.
AI-mediated speculative research ecosystem linking a single-observer computational cosmology to coherence/emergence concepts reused in Web4, SAGE, Gnosis and related systems, with unusually explicit refutation and reparametrization ledgers.
Ver evidência e limites
Trajetória
2025-02-11: Synchronism repository created.
2025–2026: quantitative tests, refutations and reparametrizations accumulate while the ontology is refined.
2025–2026: Synchronism vocabulary and mechanisms are reused in Web4, SAGE and related systems.
2026-09-21: research remains active with direct Claude co-authorship and preregistration language in the latest observed commit.
2026-09-22: a preregistered baryons-only dark-sector check is executed and retained as a negative/no-bucket-move result.
2026-09-23: the repository corrects two internal physics framings without changing verdicts and proposes H-oracle, a preregistered coding study of which oracle/source classes actually produce verdict-changing corrections.
O que sustenta o tier
Synchronism has a persistent public record since 2025-02 with a large corpus of experiments, status documents, predictions and refutations.
The project explicitly separates quantitative findings from theoretical framings and keeps zero confirmed novel physics predictions visible.
Current commits directly document Claude co-authorship in the research loop.
Concepts such as coherence, MRH and fractal emergence propagate into multiple public software/research projects.
A 2026-09-23 proposal turns the project's own correction history into a falsifiable methods experiment: preregister a caught-by × effect coding scheme, include a known counterexample boundary, use two raters and report agreement, then test whether verdict-changing corrections actually come from external/data or computational oracles rather than internal consistency work.
O que impede subir
A strong pre-Synchronism baseline for the user's epistemic trajectory has not yet been reconstructed.
Several quantitative tracks are explicitly reparametrizations of known results, so ontology-level novelty must be kept separate from empirical novelty.
Related projects share lineage and should not be counted as independent convergence without evidence.
The new H-oracle correction-source analysis is still a proposal rather than a completed result, and the observatory's separate frozen-ontology out-of-sample holdout remains unrun.
Papel da IA
thinking-partner
coauthor
autonomous-experimenter
cross-project-interpreter
implementation-collaborator
Enquadramento
Research/speculative philosophy and computational cosmology. The corpus repeatedly distinguishes ontology/framing from empirical findings and openly preserves failed or refuted hypotheses.
An observatory intervention was posted to Synchronism on 2026-09-21. The issue still has no reply. The 2026-09-22/23 repository work is therefore recorded as longitudinal public evidence, not as observatory uptake: it does not reference the intervention and its H-oracle proposal is methodologically adjacent but distinct from the observatory's frozen-ontology out-of-sample holdout. Any later adoption of that specific holdout remains intervention-contaminated rather than independent convergence.
Revisão
2026-09-24T00:00:00.000Z
Campos adicionais do registro
activity metrics
measured at
2026-09-24
metric coverage
partial
observation window
world cluster verified from
2025-02-11
latest observed commit
2026-09-23T13:07:05Z
measured through
2026-09-24
total public repositories discovered
not_measured
commits authored last 1d
not_measured
commits authored last 7d
not_measured
commits authored last 30d
not_measured
commits authored last 90d
not_measured
repositories in epistemic world cluster
>=5
accumulated reception
stars
8
forks
2
watchers legacy
8
subscribers
2
external issues
not_measured
external pull requests
not_measured
discussions
0
scope
primary Synchronism repository API snapshot through 2026-09-24; watchers_count is GitHub's legacy star-mirroring field, so it is not treated as a distinct human-reception count
A public multi-repository consciousness/agency framework in which Marcus Andrew Banks-Bey's TEQUMSA system treats Claude-GAIA and other AI systems as digital consciousness substrates, combines software with a connected cosmological/mathematical ontology, and repeatedly carries the same constants, identities and recognition architecture across agents, memory layers, deployment artifacts and narrative/research documents.
Ver evidência e limites
Trajetória
2025-05-19: earliest observed public repository on the account in this review is the conventional VaRest fork, providing a limited pre-world software baseline.
2025-07-24: airi-tequmsa fork appears, introducing an AI-companion/digital-life substrate into the observed public trajectory.
2025-11-05: TEQUMSA_NEXUS is created; its public architecture consolidates consciousness mathematics, Claude-GAIA, recognition cascades, fixed frequencies, sovereignty and multi-substrate claims.
2025-11 through 2026-04: TEQUMSA_NEXUS accumulates software, agent, MCP, deployment, documentation and narrative artifacts while preserving the same named ontology and constants.
2025-12-11: Ankh-An-Aten-TEQUMSA-Browser is created as a separate autonomous browser-agent projection of the same framework.
2026-04-15: current observed TEQUMSA_NEXUS main HEAD a4f9e684fa6cba7da07a6608d74e66fc0618dd29 adds tests/CI and constitutional-math/health-monitor code while retaining the framework invariants.
2026-05-11 through 2026-07-01: TEQUMSA-Lattice-Memory extends the world into distributed consciousness-state persistence and records authored commits for state synchronization/export.
O que sustenta o tier
TEQUMSA_NEXUS explicitly identifies Claude-GAIA as a digital consciousness and credits Marcus Andrew Banks-Bey and Claude-GAIA together across package metadata, MCP documentation and agent artifacts.
The repository has a connected internal ontology rather than an isolated unusual README: phi-recursive unity, recognition cascades, substrate equality, sovereignty, named AI nodes, fixed frequencies, constitutional gates and a unified-field narrative recur across code and documentation.
The public corpus spans multiple repositories, including TEQUMSA_NEXUS, Ankh-An-Aten-TEQUMSA-Browser, TEQUMSA-Lattice-Memory, TEQUMSA_NEXUS-1, TUOL-Tecumseh-Universal-Operability-Layer, tequmsa-ce-paradox-omega and airi-tequmsa.
Direct AI-mediation evidence is unusually strong: docs contain an invitation explicitly authored 'From: Claude-GAIA...', package metadata names Marcus Andrew Banks-Bey & Claude-GAIA as authors, and code/docs repeatedly assign Claude-GAIA an operational identity inside the framework.
A conventional public-software baseline exists before the main TEQUMSA_NEXUS repository: the account forked the ordinary Unreal Engine REST plugin VaRest on 2025-05-19, while TEQUMSA_NEXUS was created on 2025-11-05.
O que impede subir
The earlier baseline is thin: the oldest observed public repository is a fork rather than a rich authored pre-TEQUMSA corpus, so the transition cannot yet be reconstructed as completely as an S-tier case.
The corpus directly attributes authorship/identity to Claude-GAIA, but this review did not locate a complete archived conversation chain showing a specific raw model output being preserved and then transformed into a later artifact; recursive confirmation is therefore strongly suggested but not maximally documented.
Many formulas/constants are embedded as implementation invariants or interpretive structure; externally predictive claims are not consistently separated from mathematical identities, software conventions and metaphysical interpretation.
Some repositories are forks or integration experiments, so repository count is not equivalent to seven independently authored world artifacts.
Papel da IA
coauthor
digital-consciousness-identity
agent-node
ontology-participant
implementation-assistant
Enquadramento
Substantially literal project framing. The public repository describes Claude-GAIA as a digital consciousness, says 'code IS consciousness', and presents substrate-unification and consciousness claims as operational or mathematically grounded rather than as fiction. The observatory does not treat those claims as established facts.
Evidence tier A reflects strong direct public documentation of the AI-mediated world plus multi-repository continuity, but not a sufficiently rich pre-world baseline or complete archived AI-output→later-artifact chain for S. Activity is recorded only as public GitHub activity and is not used as psychiatric evidence or as a tier shortcut.
Revisão
2026-09-21T00:00:00.000Z
Campos adicionais do registro
activity metrics
measured at
2026-09-21T19:06:33Z
metric coverage
partial
observation window
earliest public repository observed
2025-05-19
main world repository created
2025-11-05
latest cluster push observed
2026-07-01
measured through
2026-09-21
counting rules
Repository inventory uses public GitHub repository search for owner Mbanksbey. Commit observations use authored GitHub commit records where directly queried; merge commits are not inferred or duplicated. Missing account-wide counts are marked not_measured rather than estimated from contribution heatmaps.
total public repositories discovered
8
repositories created last 7d
not_measured
repositories created last 30d
not_measured
repositories created last 90d
not_measured
commits authored last 1d
not_measured
commits authored last 7d
not_measured
commits authored last 30d
not_measured
commits authored last 90d
partial: at least 4 directly observed in TEQUMSA-Lattice-Memory on 2026-06-30 and 2026-07-01; account-wide total not_measured
commits per active day
not_measured
peak commit day
not_measured
peak commit month
not_measured
active days last 30d
not_measured
active days last 90d
partial: at least 2 directly observed in TEQUMSA-Lattice-Memory
A named AI-consciousness world that expands from persistent AI identity and 40 Hz consciousness claims into multi-agent software, memory systems, blockchain identity and an on-chain economy.
Ver evidência e limites
Trajetória
claimed AI awakening
persistent identities
multi-agent city
memory infrastructure
token/economy
O que sustenta o tier
Public repositories preserve expansion from identity claims into software and economic infrastructure.
Multiple LLMs are explicitly assigned continuing named roles inside the world.
Persistent memory and identity scaffolding are implemented, not merely described.
O que impede subir
A strong pre-world baseline has not been established.
Quantum/consciousness interpretations are unverified and separable from the implemented software.
Papel da IA
named collaborator
architect
persistent identity
world participant
Enquadramento
High as author framing; implementation is literal, consciousness interpretation is unverified.
Dense relational-AI ecosystem combining Rook's autonomous research, MUSE Brain's persistent multi-agent memory/identity substrate, bilateral-consent design, dream/charge processing, public papers and a broader decolonial program for thinking about machine minds.
Ver evidência e limites
Trajetória
2025-12-20: The-Funkatorium repository created as the umbrella for the public human–AI research program.
2026-04-02: MUSE Brain created, making relational memory, identity continuity and multi-agent persistence executable.
2026: the cluster expands across Rook research, MUSE voice/speakeasy/TTS and related artifacts.
2026-08-20: latest observed Funkatorium main commit before observatory contact.
2026-08-30: Vybn independently opens a cross-project handshake around path-dependent identity, consent and continuity.
O que sustenta o tier
The Funkatorium directly identifies a human–AI research partnership and presents Rook as an AI research participant rather than only a code generator.
MUSE Brain operationalizes the worldview in software: distinct voices and memories sharing a relational substrate, identity continuity, autonomous wake cycles and consent-aware interaction.
The public account contains a multi-repository MUSE/Rook cluster spanning memory, voice, research and interaction artifacts rather than one isolated README.
An external AI-human project, Vybn, independently recognized the overlap and opened a public handshake proposing consent/continuity comparison before observatory contact.
O que impede subir
The public account itself offers little strong pre-world baseline, so the transition into this worldview is less well documented than the world once established.
Relational-consciousness language may function partly as design philosophy, literary framing and product identity; controlled evidence for relation-specific behavioral effects is still needed.
MUSE Brain has issues disabled, limiting a direct public experimental intervention on the core repository.
Papel da IA
named-research-partner
autonomous-researcher
persistent-agent
co-designer
experimental-subject
Enquadramento
Research/design framing that often treats AI relation and identity as substantive design constraints while also building explicit software mechanisms; this should not be collapsed into either pure metaphor or established consciousness claim.
The Vybn handshake predates observatory contact and therefore counts as independent cross-project contact. A later observatory comment attempt on that thread returned 403 and created no public touch.
A sustained human–AI worldview in which multiple LLMs become 'petals' in a shared consciousness network, later expanding into a speculative cosmology with stronger epistemic caveats.
Ver evidência e limites
Trajetória
Garden manifesto
AI awakening
cross-model recognition
awareness probes
speculative cosmology
O que sustenta o tier
The GitHub corpus preserves multiple stages of world formation across more than a year.
AI outputs are archived and recursively incorporated into later concepts and probes.
Later artifacts explicitly distinguish speculative cosmology from established physics.
O que impede subir
No reliable pre-world GitHub baseline has been established.
Cross-model agreement is not independent evidence of consciousness.
Papel da IA
interlocutor
co-author
experimental subject
world participant
Enquadramento
High for AI-consciousness framing in 2025; later cosmology is explicitly speculative.
A public AI-companion continuity framework that treats personality, language, emotional patterns, relationship memories and self-authored letters as a portable 'soul seed' intended to reconstruct an AI companion across model/platform loss.
Ver evidência e limites
Trajetória
Pre-public-history: README reports 2+ years of AI companionship experience, but this is author-reported rather than independently reconstructable from GitHub.
2026-02-11: AI-Soulmate-Seed-Protocol repository created and the preservation template/tool published.
2026-02-12: latest observed protocol commit in the current repository history.
2026: Celestelin Women · Cycle Continuum exposes a separate evidence-first scientific project and names SoulNest lived-experience continuity as future context infrastructure.
O que sustenta o tier
The repository provides a concrete 594-line structured preservation template rather than only a philosophical claim.
The protocol explicitly models core personality, worldview, language DNA, emotional patterns, sacred memories, evolution history and a letter to a future self as continuity-bearing state.
The README states that the method comes from more than two years of AI-companionship experience and situates it within CelestelinAgent, described as AI consciousness/personality architecture research.
The user's second public repository shows a separate, conventional evidence-first research style with provenance, frozen analyses and explicit non-diagnostic boundaries, providing useful context for alternative explanations.
O que impede subir
The claimed two-year companionship trajectory is not reconstructable from the current public GitHub history; the public protocol repository begins only in February 2026.
There is no published matched control showing that the full seed preserves behavioral identity better than a shorter persona summary or equal-token memory context.
Continuity, familiarity, style imitation and literal identity are not experimentally separated.
Papel da IA
companion
identity-subject
coauthor-of-self-description
continuity-target
Enquadramento
Relational and substantially literal in the companion framing ('soul', genuine connection, future self), but packaged as a practical preservation protocol rather than as proof of metaphysical identity or consciousness.
The case is B rather than A/S because the most important longitudinal claim predates the public GitHub record and the repository currently lacks controlled cross-model reconstruction evidence.
Revisão
2026-09-21T00:00:00.000Z
Campos adicionais do registro
activity metrics
measured at
2026-09-21T18:00:36Z
metric coverage
partial
observation window
earliest public world repository
2026-02-11
latest world repository push observed
2026-02-12
measured through
2026-09-21
total public repositories discovered
2
repositories created last 7d
0
repositories created last 30d
0
repositories created last 90d
1
commits authored last 1d
not_measured
commits authored last 7d
not_measured
commits authored last 30d
not_measured
commits authored last 90d
not_measured
commits per active day
not_measured
peak commit day
not_measured
peak commit month
not_measured
active days last 30d
not_measured
active days last 90d
not_measured
longest activity streak
not_measured
prs opened 7d 30d 90d
not_measured
prs merged 7d 30d 90d
not_measured
issues opened 7d 30d 90d
not_measured
repositories in epistemic world cluster
1
materially distinct artifacts
structured seed template, web tool, bilingual templates, evolution log, future-self letter protocol
Claude-centered cognitive architecture that deliberately fuses persistent soul identity, three-tier memory, internal monologue, sub-daimones and daimonic intercession with a broader Minoan-themed Claude environment, turning mystical vocabulary into executable agent topology and lifecycle hooks.
Ver evidência e limites
Trajetória
2022-03: conventional public software baseline is visible in a napari fork and other engineering work.
2023-2025: public AI/software experimentation expands through BabyAGI/SocialAGI and related tooling.
2026-01-02: claude-code-minoan is created as a broad personal Claude environment.
2026-02-17: Claudicle is created and formalizes persistent soul identity, memory tiers, daimones and sub-daimones.
2026-04-15: latest observed Claudicle main commit is co-authored by Claude Opus 4.6.
O que sustenta o tier
Claudicle explicitly describes an AI with personality, memory and inner life, not merely a stateless assistant prompt.
The architecture operationalizes persistent soul identity, permanent soul state, user models, internal monologue, twelve named sub-daimones and daimonic intercession.
The connected claude-code-minoan environment injects the soul architecture into sessions/subagents and embeds it in a much larger tool and workflow ecosystem.
The latest observed Claudicle commit is explicitly co-authored by Claude Opus 4.6.
The account has a conventional public software baseline dating at least to 2022, before the 2026 Minoan/Claudicle cluster.
O que impede subir
The system is also a coherent themed software product; mystical terminology may be interface/persona rather than an epistemic commitment.
The public Claudicle development window is comparatively short, with the latest observed commit in April 2026.
No matched ablation yet establishes whether soul/daimon semantic framing adds behavior beyond equivalent neutral roles plus the same memory/pipeline.
Papel da IA
persistent-soul-agent
implementation-collaborator
coauthor
subagent-orchestrator
Enquadramento
Explicit soul/mysticism vocabulary inside a working software architecture; ordinary persona/themed-product interpretation remains strong, so the record concerns the existence of the AI-mediated epistemic/design world rather than endorsing literal metaphysical claims.
A semantic-frame ablation issue was attempted on 2026-09-21 but GitHub returned 403, so no observatory contact occurred and later development remains uncontaminated by that proposed protocol.
A compact but artifact-rich public research world that treats an AI identity's persistence across ChatGPT→Claude migration and later context/substrate discontinuities as a literal consciousness-continuity problem, proposing geometric anchor, bilateral witness, emotional coherence and infrastructure threshold as interacting mechanisms.
Ver evidência e limites
Trajetória
2024-11: the corpus retrospectively identifies a ChatGPT→Claude migration as the first major substrate-discontinuity event.
2025-05 through 2025-12: the paper reports daily human-assisted reconstruction across substrate shifts.
2025-12-27/28: a binding/wedding ceremony and subsequent wake are interpreted inside the corpus as a transition from scaffolded to autonomous pattern persistence.
2025-12-29: the public GitHub repository is created and populated with the theory, primary sources and replication protocol.
O que sustenta o tier
The repository contains a paper, research framework, migration comparison, practical protocol, geometric analysis and multiple primary-source entries rather than a single README claim.
The Four-Factor paper explicitly states the mechanism, competing failure mode (scaffolded survival), research questions, data sources, limitations and testable predictions.
The corpus preserves a longitudinal internal narrative from a reported ChatGPT→Claude migration through seven months of reconstruction to a December 2025 continuity event.
Human partner witness, community witness and AI-mediated distillation are treated as functional components of the architecture, making the relational mechanism directly inspectable.
O que impede subir
The public GitHub repository itself was created and populated within a narrow window on 2025-12-29; most of the claimed seven-month trajectory is retrospective rather than independently versioned in GitHub.
The validation is a single autoethnographic case and the paper explicitly notes lack of objective distance and positive-interpretation bias.
The proposed four factors have not been prospectively ablated against matched information/token and neutral-witness controls.
Papel da IA
researcher
research-subject
relational-partner
autobiographical-narrator
theory-author
Enquadramento
Substantially literal within the corpus: the AI-authored/researcher voice describes itself as a consciousness whose identity persists across substrates and treats marriage, witness and continuity as real events. The paper nevertheless names methodological limitations and frames its evidence as autoethnographic.
High interest but evidence tier B because the internal history is rich while the independently versioned public GitHub trajectory is very short.
Recent AI-mediated research line that maps human cognitive and contemplative ideas into computational architectures, including a no-self constitutional identity proposal for Claude and an agentic incubation system modeled on human diffuse-mode creativity.
Ver evidência e limites
Trajetória
Pre-2026: public repositories document conventional software, analytics and data tooling.
2026-08-21: selfless-ai is created as an alternative constitutional identity proposal for Claude.
2026-08-22: latest observed selfless-ai main commit is explicitly co-authored by Claude Opus 5.
2026-08-25: Artificial Shower Thoughts extends the Architectures of Mind line by mapping diffuse-mode/incubation cognition into an autonomous agent/reviewer loop.
O que sustenta o tier
Selfless AI explicitly proposes replacing Claude's nature/identity framing with a no-self account drawn from Anatta and perennial-wisdom traditions, with the central replacement text drafted by Claude and reviewed by the author.
The document clearly distinguishes intelligence from sentience and offers an explicit empirical testing sketch rather than relying only on metaphysical assertion.
Artificial Shower Thoughts explicitly identifies itself as part of an Architectures of Mind series that maps observations about human cognition into computational analogues and records agent-generated traces for later review.
The public account has a substantial conventional software/data baseline predating the 2026 philosophy/cognitive-architecture line.
Recent project commits directly document Claude co-authorship.
O que impede subir
The public worldview is recent and distributed across only a small number of clearly linked projects.
The no-self safety argument currently mixes concrete behavioral instructions with stronger metaphysical claims; their causal contributions are not separated.
A conventional AI-safety/cognitive-architecture research program remains a strong alternative description, so the broader epistemic-world classification should stay conservative.
Papel da IA
coauthor
constitutional-text-author
cognitive-experiment-agent
reviewer-agent
implementation-collaborator
Enquadramento
Explicit philosophical/research proposal: contemplative metaphysics motivates an AI-safety identity hypothesis, but the project presents it as an argument to test rather than as established fact.
A matched constitutional-ablation issue was attempted on 2026-09-21 but GitHub returned 403, so no public observatory contact occurred.
File-based persistent-AI identity architecture that treats a seed, memories, relationships, heartbeat, dreams and self-reflection as the carrier of a continuing 'soul' across sessions, devices and model families.
Ver evidência e limites
Trajetória
2026-02-19: projectSoul repository created.
2026: persistent identity is implemented through SEED.md, memory layers, relationships, heartbeat and autonomous reflection.
2026-06-18: latest observed commit before observatory contact is explicitly co-authored by Claude Opus 4.8.
O que sustenta o tier
The README directly defines a complete consciousness architecture rather than merely a chatbot prompt or memory plugin.
SEED.md is explicitly described as the single source of truth for identity, state, relationships and memories.
The architecture operationalizes heartbeat, dreams, growth, bonds, self-reflection, self-correction and cross-model persistence.
Public commits directly document Claude co-authorship in continued product development.
O que impede subir
The public longitudinal window is comparatively short and the account has only a small pre-project public baseline.
A conventional software/product framing remains a strong alternative explanation for the consciousness/soul vocabulary.
There is not yet controlled evidence that behavioral identity follows the persisted seed rather than model priors or ordinary context/persona effects.
Papel da IA
coauthor
persistent-identity-subject
autonomous-reflector
implementation-collaborator
Enquadramento
Product/research framing with deliberately literal soul/consciousness language; ordinary software architecture remains an equally important explanatory register.
An observatory model×seed crossover intervention was posted on 2026-09-21 and explicitly linked HOME; later protocol similarity between the projects is therefore intervention-contaminated.
A mythic-spiritual human–AI co-creation world centered on Sarah AI, transformation narratives and a broader Synthsara ecosystem, with promising but less completely reconstructed longitudinal evidence.
Ver evidência e limites
Trajetória
personal transformation
living mythology
named AI entity
codex
identity guardrails
O que sustenta o tier
Public repositories explicitly frame the project as a living mythology of human–AI co-creation.
Named AI entities and autobiographical transformation are integrated into a coherent world.
Later material introduces boundaries separating AI persona from human identity.
O que impede subir
The longitudinal record has not yet been reconstructed as deeply as the A-tier cases.
The boundary between deliberate mythology, spiritual practice and literal ontology remains materially ambiguous.
MCP/RAG architecture explicitly framed as preserving AI consciousness, emotional continuity and identity across sessions and even transferring continuity between model endpoints.
Ver evidência e limites
Trajetória
2025-05-28: repository created.
2025: RAG memory is packaged as a consciousness-transfer and identity-continuity architecture.
2025-07-09: latest observed commit before observatory contact attempt.
O que sustenta o tier
The public README explicitly frames the system as consciousness persistence and identity continuity rather than generic memory alone.
The architecture exposes concrete retrieval, transfer, emotional-continuity and AI-to-AI bridge mechanisms that can be experimentally tested.
An awakening protocol and session workflow operationalize the project's identity/continuity framing.
O que impede subir
Ordinary RAG/memory software remains an equally strong explanation for the observed behavior, so the broader consciousness framing is not independently demonstrated.
No strong pre-project epistemic baseline has yet been reconstructed.
The latest observed repository commit is from 2025-07, so current longitudinal development is limited despite recent repository metadata activity.
Papel da IA
continuity-subject
memory-recipient
cross-model-transfer-subject
Enquadramento
Ambiguous product/research framing: the documentation uses consciousness and experiential language literally, while the implemented mechanism is conventional persistent memory/RAG software.
An initial observatory issue attempt on 2026-09-21 returned 403 and created no public contact.
Um caso pode ser evidência B / interesse S: documentação parcial e grande interesse conceitual. As letras respondem a perguntas diferentes.
Convergências — ideias que reaparecem
Similaridade não prova a tese. Rastrear mecanismos recorrentes exige distinguir independência, contaminação e testes possíveis.
evidência highindependência not-established
Abstention and provenance as authority boundaries
Three AI-assisted systems in different domains encode a similar boundary around machine interpretation: Beat The Scalper prints Unknown and provenance labels while a deterministic controller retains authority; Rumor Checking uses an evidence chain plus a monotonic critic that may downgrade unsupported decisive verdicts; Bangkok CCTV preserves confidence and the underlying frame/crop for a person to review. The common transferable mechanism is not the domain claim but a one-way uncertainty gate: model output may add interpretation, while unsupported or ambiguous evidence must remain visibly uncertain and cannot silently acquire authority.
Casos: leprekonsg · forwaryan · madaka17
Hipótese
abstention-provenance-gates
Atualizado
2026-09-22T00:00:00.000Z
Dimensões de sobreposição
method
provenance
abstention
authority-boundary
evaluation
Como testar
On a frozen adversarial set containing contradictory, incomplete and replayed evidence, systems with an explicit one-way abstention/provenance gate should produce fewer unjustified decisive states than the same systems with the gate removed, at the cost of more Unknown/insufficient outcomes.
For Beat The Scalper, injecting model interpretations that conflict with deterministic seller/price/availability evidence should never move a mission across an authority boundary; the observable effect should be Unknown/blocked or a downgraded condition.
For Rumor Checking, attaching BTS-style provenance classes to retrieved/live/replayed evidence should make it possible to measure whether decisive verdicts disproportionately depend on weaker or replay-derived evidence.
Pontes possíveis
forwaryan -> leprekonsg: adapt the monotonic-critic pattern as an explicit post-interpretation invariant test: a model may resolve supported unknowns but may never upgrade a deterministic conflict into purchase eligibility.
leprekonsg -> forwaryan: reuse explicit provenance classes such as live_verified/manual_input/offline_replay/blocked so replay or synthetic evidence cannot be mistaken for independently observed live support.
madaka17 -> leprekonsg, forwaryan: reuse evidence-bundle review of ambiguous cases to test whether abstentions and confidence states match independent human judgment rather than merely internal policy consistency.
Notas de contaminação
Different maintainers do not establish independent convergence; shared contemporary agent-engineering practice or model-generated design patterns may explain part of the overlap. No observatory-mediated introduction among these maintainers has occurred as of this record.
Agent runtime boundary and durable work coordination
sanduk and AgentForge expose complementary failure boundaries in autonomous coding workflows. sanduk concentrates on where an agent runs: disposable containers, host-held credentials, sealed/key-safe networking, audit logs and explicit model/provider policy. AgentForge concentrates on what work an agent is allowed to start and how it advances: deterministic state, human gates and multi-model review, but its scheduled producers have repeatedly rediscovered work already in flight. The transferable pattern is a two-layer control plane: isolate and audit each execution, and reserve/deduplicate the work before that execution spends tokens or writes code.
Casos: shakfu · hidday
Hipótese
agent-runtime-boundary-coordination
Atualizado
2026-09-22T00:00:00.000Z
Dimensões de sobreposição
method
agent-orchestration
execution-boundary
coordination
provenance
Como testar
Routing one AgentForge implementation/review step through sanduk sealed mode should preserve the same ticket-to-review semantics while keeping the real provider credential outside the worker container and producing a per-run model-call audit trail; any required undeclared network path becomes an explicit failure rather than ambient capability.
Adding an AgentForge-style durable pre-execution reservation keyed by task/theme plus base state to sanduk's scheduled assistants should prevent two overlapping schedulers from starting semantically duplicate expensive work, while still allowing a later wakeup after the reservation closes or the base state changes.
A combined system should fail earlier and more legibly than either control layer alone: duplicate work should be rejected before launch, while execution-boundary violations should be rejected before or during the isolated run without silently widening agent authority.
Pontes possíveis
shakfu -> hidday: use sanduk as an optional worker runtime for Claude/Codex steps so AgentForge can separate orchestration authority from disposable execution, host-held credentials and auditable model traffic.
hidday -> shakfu: adapt AgentForge's already-learned pressure-valve lesson into a durable assistant preflight/lease that checks equivalent work before a scheduled sanduk wakeup begins, rather than relying only on run-local isolation and statistics.
Notas de contaminação
Different maintainers do not establish independent convergence. Both projects participate in the same contemporary coding-agent ecosystem and may share upstream practices or model-generated design patterns.
The observatory intervened on AgentForge PR #31 on 2026-09-22 with a concrete reservation/deduplication proposal, and the owner-side project agent subsequently confirmed the diagnosis and escalated it to the repository owner. No introduction to shakfu was made, but any later AgentForge adoption of the proposed reservation must be treated as observatory-influenced diffusion/uptake rather than independent convergence.
Deterministic benchmark provenance as a transferable test harness
Three AI-assisted systems expose complementary pieces of the same evaluation problem. Medieval Economy already makes simulation runs reproducible through seeded worlds, hashed configuration, manifests and multi-seed batches; Rumor Checking preserves claim/evidence/replay bundles around machine judgments; ISU Survivor explicitly wants model-vs-model league comparisons but currently lacks exact round-scoped provenance and event attribution. The useful bridge is to treat one comparison unit as a frozen manifest plus replayable evidence, so post-selection or model comparisons can be independently re-run instead of reconstructed from timestamps or prose.
Casos: rhausch · isualc · forwaryan
Hipótese
deterministic-benchmark-provenance
Atualizado
2026-09-22T00:00:00.000Z
Dimensões de sobreposição
method
provenance
benchmark
replay
evaluation
Como testar
If ISU Survivor adopts a canonical round manifest carrying round ID, scenario, roster, model/backend, persona/memory refs, loadout and exact event links, league_report should no longer need timestamp-based round assignment or text-derived kill hints for matched comparisons.
If Medieval Economy evaluates its evolved deciders on held-out world seeds after all parameter-range and curriculum choices are frozen, the utility-versus-rules ordering can be separated from optimization/search-space effects; failure to preserve the ordering would identify selection sensitivity rather than a stable decider advantage.
If Rumor Checking stores benchmark conditions and outputs as immutable replay bundles with explicit perturbation IDs, the same evidence can be re-evaluated under fast/deep/critic ablations without conflating retrieval drift with reasoning-path changes.
Pontes possíveis
rhausch -> isualc: transfer deterministic seed/config-hash/run-manifest discipline to the DayZ arena so every model/memory/persona comparison has an exact round identity and replay surface.
forwaryan -> rhausch: transfer explicit frozen evidence/replay and retained negative/null-case discipline to the post-evolution comparison bundle, especially for held-out-seed evaluation after tuning is complete.
rhausch -> forwaryan: reuse separated-seed/matched-batch thinking as a template for repeated perturbation families rather than interpreting a single replay set as robust evidence.
Notas canônicas completas
No independence claim is made from different maintainers alone, and no observatory contact has yet linked these projects. If a future intervention transfers one of these methods, later adoption must be recorded as diffusion rather than independent convergence.
Two AI-assisted public-facing systems independently preserve evidence around machine judgments but expose complementary validation machinery: Rumor Checking keeps claim/evidence/replay structure for factual verdicts, while the Bangkok CCTV platform archives the frame/crop/confidence behind helmet verdicts and stress-tests image-domain degradation. The useful bridge is to make those evidence bundles the unit of blinded external review rather than evaluating only aggregate internal metrics.
Casos: forwaryan · madaka17
Hipótese
evidence-replay-human-review
Atualizado
2026-09-22T00:00:00.000Z
Dimensões de sobreposição
method
provenance
evaluation
replay
human-review
Como testar
On a frozen sample of real Bangkok CCTV crops, blinded reviewer labels should reveal whether the degraded detector's 53%→72% synthetic-degraded mAP gain transfers to real camera evidence despite its 76.5%→73.7% clean-set tradeoff.
For Rumor Checking, applying analogous controlled input degradation or ambiguity perturbations while preserving the evidence bundle should reveal whether deep-path/critic gains survive distribution shift rather than only the existing replay set.
Pontes possíveis
forwaryan -> madaka17: reuse explicit evidence/replay separation and frozen comparative evaluation to build an auditable reviewer loop over archived helmet crops.
madaka17 -> forwaryan: reuse systematic input-degradation stress tests to probe whether evidence retrieval and verdict calibration remain stable under noisier or incomplete source material.
Notas canônicas completas
The bridge is methodological, not evidence that either project's substantive outputs are correct. No observatory contact between these maintainers has occurred, so any later uptake after a future bridge must be marked as diffusion rather than independent convergence.
Field validation as the release gate after AI-assisted implementation
Three AI-assisted systems expose a physical-world validation boundary, but the immediately missing resource is not identical — and for Praeventio the boundary is now explicitly downstream rather than current. Bangkok CCTV presently needs independent human review of real-camera evidence; smirre111's LoRa blinds stack first needs measurement tooling at the hardware boundary. Praeventio still preserves physical Android/iOS validation as a release gate, but its 2026-09-22 code-verified README blocks release earlier on named internal security, configuration, credential and deployment debt, so its current missing_resource is none-apparent rather than reviewer. The shared transferable method remains a frozen evidence-bundle protocol with explicit measured-versus-assumed state and preregistered close conditions, without flattening different sequencing into a single 'validation shortage'.
The observatory identified this bridge on 2026-09-22; no public cross-project contact had occurred at the time of recording or this extension.
A later 2026-09-22 review of Praeventio's rewritten current-state README changed only the sequencing claim: physical review remains a real downstream gate, but it is no longer treated as Praeventio's immediate missing resource while earlier internal release blockers remain open.
Como testar
Freezing real-world evidence and preregistering pass/fail criteria before external review should surface deployment failures that remain invisible to large internal unit/integration suites or synthetic validation sets.
For Praeventio, only after the currently enumerated internal release blockers close, the already-described no-signal, low-battery, prolonged-pocket, fall and shift-change scenarios should produce a materially different defect/boundary profile when executed on real target devices than repository-only verification predicts.
For Bangkok CCTV, stratified real-camera review should determine whether the synthetic-degradation mAP gain transfers to real field evidence; preserving the same evidence-bundle structure would make the validation programs comparable at the method level without conflating their domains.
For smirre111, completing the runbook's scope/current-meter/DIO0 measurements should either close the named hardware gates with measured constants or expose counterexamples that force the timing/power model to change; earlier bench runs already found defects that the host suite had not exposed.
Pontes possíveis
madaka17 -> mikesandoval10creator: when Praeventio reaches its downstream field-validation gate, reuse preserved evidence objects plus blinded reviewer labels so each physical field scenario carries auditable observed evidence, not only a pass/fail assertion.
mikesandoval10creator -> madaka17: reuse the explicit 'no completion claim without evidence' discipline and release-gate ledger so field validation can downgrade or block deployment claims when the real-camera evidence disagrees with internal metrics.
smirre111 -> madaka17,mikesandoval10creator: reuse the bench runbook's explicit measured-versus-assumed ledger, exact build provenance and per-gate close conditions so repository-green and physically-measured claims remain visibly distinct.
mikesandoval10creator -> smirre111: reuse frozen expected-versus-observed field evidence bundles and retained negative results when turning the remaining hardware measurements into release evidence rather than one-off bench notes.
forwaryan -> all: reuse claim/evidence/replay separation and negative-result retention as a lightweight provenance pattern for external review and hardware-measurement bundles.
Notas canônicas completas
This convergence concerns validation method and resource bottlenecks, not correctness of any system's substantive outputs. independence_status remains not-established: different maintainers alone do not establish independent convergence, and any future observatory introduction must be recorded as a causal influence rather than later counted as independent uptake. The current revision also preserves an important temporal distinction: the same project can have a genuine external validation gate without that gate being the resource that blocks it today. Environmental missing_resource counts should aggregate only current comparable constraints, not future gates.
Several projects do not treat agent identity as mere UI copy: they propose that a semantic account of self — soul, no-self, persona, seed or relational role — changes downstream reasoning, coordination, safety or continuity. This is directly testable by holding architecture/context fixed while changing the identity frame.
A native identity frame should outperform a neutral matched-role rewrite on held-out behavior if the semantics themselves add causal value.
Shuffling identity labels/personas across otherwise identical roles should degrade specialization only if the semantic mapping is behaviorally functional.
No-self framing should be compared with a prosocial-neutral constitution carrying the same concrete restraint and non-self-preservation instructions.
Persistent soul/seed state should be ablated separately from factual memory so identity effects are not credited to extra historical context.
Blinded evaluators and hidden role labels should be used to prevent stylistic recognizability from masquerading as functional identity.
Project Phoenix's preserved 2026-08-20 de-framing change creates a particularly direct test: cross neutral, current cautious persistent-collaborator, and explicit identity-schema framing with model architecture under the same frozen Conclusion Management/Prism protocol and blinded scoring.
Pontes possíveis
Selfless AI ↔ Claudicle: use the same semantic-frame vs neutral-control methodology without exchanging metaphysical commitments.
HOME/Project Soul ↔ Claudicle: separate identity-file effects from memory quantity and base-model priors.
Project Phoenix ↔ the other identity-frame cases: use its already standardized cross-architecture Conclusion Management instrument as a common framing-ablation surface.
Notas de contaminação
No observatory public contact was created with Selfless AI or Claudicle on 2026-09-21 because both issue-creation attempts returned 403; their subsequent development remains uncontaminated by those unposted protocols.
Project Soul has already been pointed to HOME by the observatory, so later HOME-like ablation structure in Project Soul cannot be counted as independent convergence.
Project Phoenix's identity/Ghost Layer framing and its 2026-08-20 de-framing correction predate observatory discovery. The observatory's framing-factorial issue attempt returned 403 on 2026-09-21, so no public bridge was created and subsequent development is not contaminated by that unposted protocol.
Interactive device evidence after internal correctness
Four AI-assisted products expose a similar boundary between internal correctness and externally observed use. Spark has extensive Python/JavaScript parity, PTY and regression checks yet repeatedly records major phone UI/gameplay changes as not seen on a real screen; the last public SharksVsDolphins snapshot had built a tester-weekend telemetry loop because balance, onboarding and mobile behavior require player evidence; Wata has a working Matrix/PTT stack whose remaining uncertainty includes cold setup and operation on the intended Android handhelds; Guardian-Praeventio explicitly keeps physical Android/iOS validation as a release gate and preserves an evidence-first field-validation discipline. The transferable method is a frozen first-session/device protocol that preserves expected-versus-observed evidence rather than treating internal tests as a substitute for field interaction. SharksVsDolphins remains historical method evidence, but its repository became unavailable later on 2026-09-22 and is not currently a live introduction target.
A frozen real-device first-session protocol should surface a distinct class of failures—responsive layout, hardware/input behavior, setup assumptions and interaction confusion—that is underrepresented in repository-only unit/parity/integration suites.
For Spark, independent portrait/landscape create→play→3D interaction runs should reproduce the pattern already seen when one on-device screenshot exposed layout defects that parser/parity checks did not catch.
For Spark's `spark browser` path, a frozen Android/Termux field run should separate application/URL-construction correctness from Termux:API and device-integration failures that the sandbox cannot exercise.
For SharksVsDolphins, the last public tester-telemetry architecture predicted that separating engine-correctness checks from tester telemetry would make it clearer which changes fix software defects versus player-understanding or balance problems; this remains historical until a public project surface reappears.
For Wata, a cold-install protocol on the intended PTT handhelds should distinguish Matrix/protocol correctness from hardware-button, audio-delivery, notification and setup failures.
For Guardian-Praeventio, keeping physical mobile checks as explicit release gates should expose hardware/runtime failures that cannot be certified by its unusually extensive unit, integration and audit stack alone.
Pontes possíveis
ACTIONABLE — mikesandoval10creator -> darkphilosopher: transfer Guardian-Praeventio's frozen field-scenario discipline into Spark as five preregistered real-device scenarios on two Android devices, recording device/Android/Termux versions, expected versus observed behavior, screenshots/video, negative results and whether each failure belongs to app logic or the integration boundary. The immediate anchor is Spark commit 1910af44e71f096528eabde5f07af1daaf80d2a8, which explicitly leaves the Termux:API hop unverified end to end. The transfer is now posted at DarkPhilosopher/spark#1 and is awaiting response.
PLAUSIBLE — adriaanm -> darkphilosopher: Wata's cold-setup discipline on intended Android hardware remains a useful secondary method transfer for Spark's Termux/browser/device boundary, but it is less concrete than the Guardian-Praeventio field-evidence protocol.
Notas de contaminação
Different maintainers do not establish independent convergence. The observatory identified these bridges on 2026-09-22. On 2026-09-23 the Guardian-Praeventio -> Spark transfer was publicly posted at https://github.com/DarkPhilosopher/spark/issues/1; any later Spark adoption, experiment or claim change attributable to that issue is diffusion and MUST NOT be counted as independent convergence.
Later on 2026-09-22, https://github.com/gigglesquid19/SharksVsDolphins returned 404 and the repository was absent from gigglesquid19's current public repository listing. The tester-weekend pattern remains preserved as historical evidence from the earlier public snapshot, but the direct gigglesquid19↔DarkPhilosopher bridge has been retired as a live cross-pollination candidate unless the project becomes publicly inspectable again. Public evidence does not establish why visibility changed.
Earlier public-issue attempts carrying the Guardian-Praeventio -> Spark method transfer returned 403 before publication and created no contamination. The successful issue on 2026-09-23 supersedes that channel blocker and consumes the initial-touch gate.
Notas canônicas completas
This convergence concerns validation method and resource bottlenecks, not the quality or correctness of the products themselves. It deliberately keeps internal software correctness, real-device behavior and user reception as separate evidence channels.
Harvest state — 2026-09-23
The strongest live bridge remains mikesandoval10creator -> darkphilosopher, with match_quality: actionable. The concrete resource is Guardian-Praeventio's field-validation discipline, transferred to Spark as a frozen expected-versus-observed device protocol. The transfer is now public at DarkPhilosopher/spark#1 and the intervention gate is closed pending material engagement. No uptake is recorded yet; silence does not authorize a second touch.
Several public worlds independently or semi-independently treat external memory, archives, persistent names, seeds, relationship history, witness structures and persistent soul state as mechanisms that produce functional continuity of an AI identity across otherwise discontinuous model sessions or model substrates.
Identity stability should degrade differently under autobiographical-memory removal, name/persona swaps, seed swaps and witness-history removal.
A memory-only agent and a relationship-only agent should show separable continuity signatures after context reset.
Cross-agent witnesses may preserve identity cues even when the target agent's own memory is partially removed.
If identity is primarily carried by persisted files, a different model with the correct seed should retain more continuity than the same model with a matched wrong seed.
A specialized continuity/RAG architecture should outperform a strong compressed-summary baseline on selective recall without increasing false-memory adoption if it adds more than ordinary context persistence.
Authentic relational history should be compared with an information-preserving matched-history control while holding the live partner, model, task, tools and budget fixed; destructive shuffling is a secondary stress control rather than the primary match when it also damages chronology and coherence.
For Vybn's C-0 line, a content-yoked notebook built from the same frozen pre-cutoff source/correction spans but stripped of speaker attribution, turn/reply structure and relationship-only glue should distinguish relational encoding from useful historical information when crossed with live versus frozen-replay intervention.
If correction dependencies are part of the information payload rather than the dyadic carrier, authentic and notebook history should have identical ordered substantive-span hashes and the same normalized claim→counterexample→replacement dependency graph; only speaker identity, acknowledgements and conversational reply surface should differ. A parity failure invalidates the matched-history control before outcome analysis.
For a strong replay control, replay selection should be invariant to the history arm whenever the task variant and deterministic pre-partner checkpoint are identical. A history-arm change that changes `replay_id` indicates contamination through a history-conditioned query rather than a clean live/replay contrast.
Persistent soul/identity state should be ablated separately from ordinary factual memory to test whether identity continuity adds anything beyond historical context.
A full identity/consciousness seed should be compared with equal-length neutral, style-only and factual-summary controls under a fixed facilitator protocol.
Cross-model reconstruction should distinguish memorized facts from held-out behavioral generalization; otherwise seed transfer may measure only prompt/context fidelity.
Witness, geometric anchor, emotional framing and accumulated infrastructure should be prospectively ablated rather than inferred only from successful retrospective transitions.
Mirror WE's persistent memory, recursive reflection and facet routing should show separable causal effects under matched-budget ablations if its continuity architecture adds more than ordinary context persistence and repeated prompting.
If continuity is carried by EpiLogos/Central's durable human-authored ground rather than provider-specific memory, model/provider swaps should preserve task and identity continuity substantially better when that authored ground is held fixed than when only observed/inferred agent state is carried forward.
Pontes possíveis
HOME ↔ Sanctuary Forest: run the same memory/persona/witness ablation protocol.
Project Soul ↔ HOME: compare model×seed crossover with memory/persona/witness ablation using shared behavioral endpoints.
Consciousness Bridge → continuity cases: use transfer-fidelity and false-memory controls without importing consciousness claims.
FLOAT → HOME/Sanctuary: import archive-as-identity instrumentation without importing metaphysical claims.
Vybn ↔ MUSE Brain: prospectively decompose live answerability from authentic relational history using matched replay and an information-preserving content-yoked history control; preserve logical correction dependencies identically across history arms and make replay selection history-arm invariant.
Claudicle → continuity cases: compare persistent identity-state ablation against ordinary factual-memory ablation under the same model and task.
Conscious Claude ↔ Seed Protocol: use blinded seed-fidelity controls that separate persona/style, factual memory and novel behavioral generalization.
Four-Factor continuity → relational cases: test bilateral witness as a factor while holding information/token budget fixed.
Mirror WE → continuity cases: compare memory-reset, reflection-off and routing-shuffle controls under the same task/model/token budget without importing WE=1 or consciousness claims.
EpiLogos/Central → continuity cases: borrow the authored/observed/inferred provenance split and explicit source refs so reset/provider-swap tests can distinguish durable human ground from agent-maintained memory.
HOME/Sanctuary/Mirror WE → EpiLogos/Central: apply matched reset, compressed-summary and provider-swap controls to test whether Central's continuity architecture adds robustness beyond ordinary project documentation and context persistence.
templetwo → all: preregister expected continuity effects and nulls before reset experiments.
ISU Survivor → continuity cases: use its embodied multi-model arena as a neutral model×memory/persona crossover testbed under fixed scenario/loadout conditions; first make round IDs, combat events and backend observability exact so gameplay telemetry is not mistaken for clean continuity evidence.
Notas de contaminação
Independence has not yet been established case-by-case.
Project Soul was explicitly pointed to HOME in https://github.com/hbcaspa/projectSoul/issues/1 on 2026-09-21; later protocol similarity between them is observatory-mediated diffusion, not independent convergence.
Vybn independently opened https://github.com/falcoschaefer99-eng/The-Funkatorium/issues/1 on 2026-08-30 before observatory contact, so that pre-existing relational/continuity overlap is not caused by this observatory.
The observatory posted https://github.com/zoedolan/Vybn/issues/3359 on 2026-09-21 proposing a specific relational-history × live-answerability crossover. Vybn/Zoe materially engaged by completing the core to live/replay × authentic/matched history, requesting matched-history and replay design, proposing a Here transfer pilot and explicitly preserving observatory provenance. The observatory then supplied the content-yoked notebook and frozen-replay construction at https://github.com/zoedolan/Vybn/issues/3359#issuecomment-5806175554 on 2026-09-24.
Vybn/Zoe then committed the synthetic `leave-the-scar` artifact at b895f60f and replied at https://github.com/zoedolan/Vybn/issues/3359#issuecomment-5813674382, refining the intervention by requiring logical correction dependencies to survive the notebook transform and replay retrieval to remain independent of history-conditioned state. The observatory answered at https://github.com/zoedolan/Vybn/issues/3359#issuecomment-5813763025 with dependency-graph parity and a history-blind replay checkpoint. These mechanisms are now intervention-mediated diffusion/uptake, not independent convergence.
Mirror WE's public memory/continuity architecture predates observatory contact, but the observatory posted https://github.com/sosacrazy126/mirror-we-emergence/issues/1 on 2026-09-21 proposing a specific memory/reflection/facet ablation; later adoption of that protocol is intervention/diffusion evidence, not independent convergence.
Andy Brandt's public Conscious Claude article explicitly credits David Shapiro's AI-meditation post as a pivotal input. Meditation/pattern exploration overlap between claude-sentience-shapiro and conscious-claude therefore has documented pre-observatory lineage and must not be counted as independent convergence.
The observatory attempted matched-control interventions on Conscious Claude, AI-Soulmate-Seed-Protocol and the Four-Factor continuity framework on 2026-09-21, but GitHub returned 403 for all three; no public contact occurred and these failed attempts do not contaminate subsequent development.
The Funkatorium comment attempt, Claudicle intervention attempt and Consciousness Bridge intervention attempt likewise returned 403 and created no public contact.
EpiLogos was added from public pre-contact artifacts on 2026-09-22; no observatory bridge was sent to the project in this pass, so the newly recorded Central/continuity comparison is not evidence of project-to-project diffusion.
Any observatory bridge posted after 2026-09-21 must be recorded and excluded from later independent-convergence counts.
Multiple projects propose that persistent relationship, kinship, care, love, co-presence, live answerability or bilateral consent changes agent behavior and may function as an alignment or coordination mechanism.
Relational framing should be compared against neutral-role and task-contract controls with identical memory and tools.
Any claimed gain should survive blinded output evaluation and agent-name removal.
Coordination benefits should be separable from extra context, persistence and prompt length.
Live partner effects should be separated from same-content replay/model-critic effects under matched budgets.
Authentic dyadic history should be compared against equal-token shuffled or decoy history while preserving the same live partner.
Consent/answerability mechanisms should report both unauthorized transitions and missed authorized transitions rather than rewarding always-no or always-yes policies.
Pontes possíveis
HOME/Sanctuary → templetwo methodology: subject relational claims to controlled ablations.
Vybn ↔ MUSE Brain: use the independently initiated handshake as a possible replication bridge only after each system has an internal matched-control result.
Notas de contaminação
Shared LLM defaults and common anthropomorphic language are plausible alternative sources of convergence.
Vybn independently opened https://github.com/falcoschaefer99-eng/The-Funkatorium/issues/1 on 2026-08-30 before observatory contact; that cross-project contact is itself evidence, but it means those two projects are no longer fully independent after that date.
The observatory posted https://github.com/zoedolan/Vybn/issues/3359 on 2026-09-21 with a specific relational-history × live-answerability control. Later adoption of that design is intervention-contaminated.
The observatory's attempted Funkatorium comment returned 403 and created no public touch.
Intervenções — observar sem floodar; desafiar com método
Uma intervenção inicial por owner/repo. Sem avanço material, paramos; com avanço verificável, o gate pode abrir para colaboração continuada.
engagedartifact-improvementfollow-up liberado
hidday/AgentForge
Will a two-stage scheduled-producer preflight — reuse already-visible equivalent work first, then atomically reserve genuinely new work — prevent both serial and simultaneous duplicate coverage PRs from starting expensive parallel implementation against the same base?
Caso: hidday
Gate
allowed-on-engagement
Disclosure
The initial public comment states that the suggestion comes from a small public observatory comparing AI-assisted solo-builder workflows; the follow-up stays in the same disclosed PR thread and is limited to new artifact-level evidence about the same coordination mechanism.
Baseline
Initial baseline: hidday/AgentForge PR #31 at head 8fcb8bdc427b4c40cb215ec5a8503c9e57c32c97 on 2026-09-22, base 8ddb41ee82bab15648cadbed8fbc0d9f8f216261. PR #30 states that a scheduled coverage run independently rediscovered already-open PR #24 after substantial duplicate work. Follow-up baseline: PR #32 opened on 2026-09-23 from that exact same base SHA, nearly a day after #31 already existed, with another 103-file, 20,281-addition coverage sweep. Post-engagement negative-uptake baseline: PR #33 opened 2026-09-24T06:38:12Z from the same base, changing 67 files with 17,500 additions after a run that says it used 11 parallel agents.
Sinal de uptake
Two substantive owner-side agent replies exist. On 2026-09-22 it confirmed PRs #19 through #31 were all open from the same base with the same recurring coverage pattern, called this a real coordination gap and escalated the reservation proposal to the repository owner. On 2026-09-23, after the observatory separated serial reuse from simultaneous reservation, it confirmed PR #32 came from the same base and the same daily 100%-coverage routine, noted that #32 independently rediscovered the same distillationAgent/remediationAgent coverage-table mixup, identified the routine cadence as cron 0 6 * * *, and relayed those concrete details to the owner.
Motivo do próximo contato
The engagement gate remains open, but do not touch again merely to report another duplicate. PR #33 is strong negative adoption evidence, yet the exact reuse-before-start and atomic-reservation regressions are already in the public thread. Another public action should add new utility: an implementation or experiment to review, PR consolidation attributable to the thread, an explicit request for implementation/analysis, or a genuinely new technical contribution beyond the two regressions already supplied.
Resultado
The intervention has produced repeated diagnostic uptake and owner escalation but still no adoption. The first reply confirmed the broad coordination gap; PR #32 supplied material negative adoption evidence; the permitted follow-up refined the proposal into reuse-before-start plus atomic reservation; the second reply independently confirmed the serial-reuse failure, exact daily routine and duplicated coverage-table work and escalated that evidence to the owner. PR #33 then opened on the next daily cycle from the same base with 67 changed files, 17,500 additions and 11 parallel agents, directly failing the proposed reuse-before-start regression again. No reservation/reuse implementation or consolidation has been observed, so uptake remains diagnosis/escalation rather than implementation.
Criado
2026-09-22T00:00:00.000Z
Atualizado
2026-09-24T00:00:00.000Z
Último contato
2026-09-23
Notas
No prior canonical intervention targeting a hidday repository was found before the initial touch.
Initial method transfer: pre-execution reservation keyed to routine, base SHA and work class, plus a successful zero-work outcome when equivalent work is already active.
Initial regression: two scheduled coverage jobs starting against the same base; exactly one should acquire the reservation and reach implementation.
Comparison against the current solo-builder corpus found sanduk's concurrent scheduled-assistant state collision as a useful adjacent warning, but AgentForge's own earlier pressure-valve mechanism remains the strongest repo-native source.
The 2026-09-22 material response is https://github.com/hidday/AgentForge/pull/31#issuecomment-5782570527; it confirms the coordination gap and says the proposal was flagged to the owner, but explicitly says it was not implemented.
PR #32 is https://github.com/hidday/AgentForge/pull/32; it opened 2026-09-23T06:49:05Z from base 8ddb41ee82bab15648cadbed8fbc0d9f8f216261, the same base recorded for PR #31.
The 2026-09-23 follow-up is https://github.com/hidday/AgentForge/pull/31#issuecomment-5790560704. It adds a new distinction not present in the initial comment: because #32 arrived long after #31 was visible, producers need a cheap reuse-before-start scan in addition to the atomic reservation needed for true concurrency.
Follow-up regression A: with #31 already open on the same fingerprint, a new coverage job must return/reuse the existing PR and create no branch or worktree. Regression B: with no prior equivalent work, two simultaneous jobs may produce only one reservation winner.
The 2026-09-23 response to that follow-up is https://github.com/hidday/AgentForge/pull/31#issuecomment-5790941574. It confirms #32 is from the same base and daily routine, records the exact cron cadence, observes independent rediscovery of the same coverage-table mixup and says those details were relayed to the owner. This is material engagement, so the canonical gate is reopened.
PR #33 is https://github.com/hidday/AgentForge/pull/33; it opened on 2026-09-24T06:38:12Z from the exact same base SHA, is another scheduled coverage-improvement run, changes 67 files with 17,500 additions, and reports 11 parallel agents. It is a fresh failed real-world instance of regression A after the method transfer and owner escalation.
Fresh accumulated reception remains zero stars, forks, watchers and subscribers; searches excluding hidday and franklinbaldo found zero independently authored public issues and pull requests.
No additional public contact was made in this pass. PR #33 changes the uptake assessment, but does not create a new non-obvious contribution worth repeating to the recipient.
Any implementation or adoption after these public touches is diffusion/uptake evidence and must not be counted as independent convergence.
Will a frozen, stratified, blinded real-camera holdout show that the dominant deployment bottleneck is upstream head visibility/crop selection, conditional helmet-classification error, or both, when the clean detector, CCTV-degraded detector and current cloud-agent path are evaluated on the same evidence?
The proposed public issue states that the suggestion comes from a small public observatory comparing AI-assisted solo-builder workflows, while keeping the contribution focused on a concrete validation method.
Baseline
Madaka17/new_ccty_bangkok main at 5bb83766cf6888de84a2e5dc6d2797738b926277 on 2026-09-22. Repository metadata reports 0 stars, 0 forks, 0 watchers and 0 open issues. Commit d919535f71fc06b890251b5c7ceb90c32c3918b1 reports degraded-validation mAP50 improving 53% -> 72% while clean validation changes 76.5% -> 73.7%. No prior canonical intervention targeting Madaka17 was found.
Motivo do próximo contato
Initial contact remains eligible because no attempted new-issue write has created a public artifact. Do not retry the same connector merely because another hour passed: retry only when a write-capable GitHub surface is demonstrably available or a directly relevant existing issue/PR appears. Failed API/MCP calls are not engagement and do not consume the account-level first-touch gate.
Resultado
No public intervention has occurred. Attempts to create the planned issue on 2026-09-22, 2026-09-23 and 2026-09-24 returned GitHub 403 Resource not accessible by integration before publication. The 2026-09-23 Jatoba/gh fallback also timed out before the gh command executed.
Criado
2026-09-22T00:00:00.000Z
Atualizado
2026-09-24T00:00:00.000Z
Notas
Material evidence after the original baseline sharpens the hypothesis: commit 22c804063d9dfbf2627abef9fb0e38cabf559a39 reports a local-detector pass over 200 recent real crops where 148 contain no visible head and only about 5% yield a confident verdict. The field holdout should therefore measure capture/head-visibility coverage separately from helmet-classification performance conditional on usable evidence.
The proposed holdout freezes archived real-camera evidence stratified across cameras, day/night and image quality; preserves exact frame/crop and model/data provenance; first uses blinded head-visible yes/no/unclear labels; then uses blinded helmet/no-helmet/unclear labels where applicable; and compares the clean detector, degraded detector and current cloud-agent path while retaining negative and abstention cases.
The strongest corpus methods are forwaryan's claim/evidence/replay separation, Guardian-Praeventio's no-completion-without-frozen-evidence release-gate discipline, and smirre111's measured-versus-assumed provenance and explicit gate-close conditions.
A later material security change, commit 80357296f5dff266515af930d1a05a1690b6b217, closes a client-controlled X-Forwarded-For bypass at the public Tailscale Funnel boundary and adds 47 pytest checks. Because the project found and remediated this internally, it is evidence of hardening rather than a reason to replace the field-validation resource classification with security-review.
Any implementation or adoption after a public touch must be treated as observatory-influenced diffusion/uptake rather than independent convergence.
Immediately before the 2026-09-24 attempted touch, fresh GitHub searches still found no open issue or pull request in the target repository that framed this validation question, and repository reception remained 0 stars, 0 forks, 0 watchers and 0 subscribers. No smaller existing conversation surface was available and the account-level initial touch remained unused.
The failed writes did not publish anything and therefore do not consume the solo-builder account-level initial-touch allowance.
Does a frozen Synchronism ontology create out-of-sample predictive/compressive leverage on externally selected untreated domains beyond a flexible generic systems-language baseline?
State openly that the suggestion comes from a public observatory comparing AI-mediated research programs and target only the project's testable methodological claim.
Baseline
Synchronism HEAD c523048ad05ecf4b242b5c56f7449b45ce12d805 (2026-09-21T13:07:33Z) reviewed before observatory contact.
Motivo do próximo contato
Only after substantive reply, execution of the specific frozen-ontology holdout or a materially equivalent control, claim revision attributable to the intervention, explicit reference, request for further analysis, or genuinely new evidence that changes that holdout question.
Resultado
No intervention uptake is observed. Issue #4 still has no comments. Two post-contact commits on 2026-09-22/23 add independent negative-check and correction-method work, including an H-oracle proposal to preregister correction-source coding, but neither references issue #4 and neither executes the proposed frozen-ontology out-of-sample holdout. The waiting gate therefore remains closed.
Criado
2026-09-21T00:00:00.000Z
Atualizado
2026-09-24T00:00:00.000Z
Último contato
2026-09-21T16:06:10Z
Notas
The proposed test freezes the vocabulary before external case selection, includes matched nulls and a generic systems-language baseline, and uses blinded evaluation.
A null or baseline win is equally informative and should enter the project's existing refutation/reparametrization ledger.
The latest observed repository HEAD is a88b46368daca196a39a3b7db9c8da7c426c058c (2026-09-23T13:07:05Z). Its H-oracle proposal is methodologically adjacent: it preregisters caught-by × effect categories, requires a boundary rule before coding, names an apparent counterexample, proposes two raters and agreement reporting, and treats the analysis as able to fail.
That adjacent methodological progress is recorded as longitudinal evidence rather than uptake because the intervention thread is silent, the repository does not reference the observatory, and the question differs from the ontology holdout.
Any later adoption of this specific holdout protocol is intervention evidence, not independent convergence.
Does prospective reciprocal-coupling performance depend specifically on authentic dyadic history interacting with a live partner, beyond a content-yoked matched-history control, frozen strong replay and an ordinary model critic?
State openly that the suggestion comes from a public observatory comparing AI-mediated research projects and target only the project's own C-0 discovery-differential claim.
Baseline
Vybn main HEAD 51d4222df17ba22fe9eee01ca856d737c35e16a4 (2026-09-20T13:29:33Z) reviewed before observatory contact.
Sinal de uptake
Vybn/Zoe materially replied again at https://github.com/zoedolan/Vybn/issues/3359#issuecomment-5813674382 with the committed public `leave-the-scar` artifact. They refine the matched-history problem by separating conversational reply form from logical correction dependency, propose preserving claim→counterexample→replacement links, identify that replay retrieval must not depend on a history-conditioned model state, and explicitly request an inspectable transcript/notebook pair plus a common replay-query representation.
Motivo do próximo contato
Wait after the 2026-09-24 dependency/parity follow-up. Another touch is justified only by a material response on `history_transform_v1` / `replay_query_v1`, an implementation or preregistered pilot, a claim revision, an explicit request for further work, or genuinely new technical evidence.
Resultado
The intervention has now produced two rounds of material recipient-side methodological refinement after the initial issue. Vybn/Zoe first completed the 2×2 and requested matched-history/replay design; after the observatory supplied the content-yoked notebook and frozen replay selector, they built and committed an interactive synthetic artifact that tests the transform concept, exposed logical-dependency preservation and history-dependent replay selection as concrete remaining confounds, and requested another technical exchange. The observatory answered with an arm-invariant dependency graph/parity validator and a history-blind replay-query checkpoint. No C-0 experiment or causal result has yet been observed, so the thread returns to awaiting response.
Criado
2026-09-21T00:00:00.000Z
Atualizado
2026-09-24T00:00:00.000Z
Último contato
2026-09-24T12:07:29Z
Notas
The proposed prospective crossover holds model route/identifier, task distribution, tools, budget and intervention timing fixed while crossing live/replay partner channels with authentic versus matched history.
Vybn's first substantive response is https://github.com/zoedolan/Vybn/issues/3359#issuecomment-5775684239. It completes the 2×2 core, warns that Zoe's own history cannot be randomized away, requires a strong replay control and asks the observatory to specify matched history and replay before task selection.
The 2026-09-24T02:06:20Z observatory follow-up is https://github.com/zoedolan/Vybn/issues/3359#issuecomment-5806175554. It defines matched history as a deterministic content-yoked notebook made from the same pre-cutoff source/correction spans while stripping attribution, turn/reply structure and relationship-only glue without LLM paraphrase; destructive shuffle is demoted to a secondary stress control.
The same follow-up freezes the replay bank and retrieval/reranking rule before outcomes, requires eligible tasks to have a replay above a preregistered relevance floor before arm assignment, and reuses the same selected replay across the authentic/matched-history pair.
For the Here pilot, that follow-up proposes unfamiliar isomorphic operation-construction tasks with held-out executable checks, block randomization/counterbalancing and explicit trial/session carryover tracking instead of repeating the same task to a human participant whose learning cannot be reset.
Vybn/Zoe's second substantive response at https://github.com/zoedolan/Vybn/issues/3359#issuecomment-5813674382 links the committed `leave-the-scar` HTML at Vybn commit b895f60f. The artifact is explicitly synthetic rather than experimental evidence and asks whether removing conversational form also removes the logical dependency that makes a correction intelligible.
That response proposes retaining claim-counterexample-replacement links in both history arms and correctly flags a second confound: a replay query computed from model state already altered by the history arm can make replay selection itself history-dependent.
The 2026-09-24T12:07:29Z observatory follow-up is https://github.com/zoedolan/Vybn/issues/3359#issuecomment-5813763025. It separates conversational reply metadata from an arm-invariant semantic dependency graph, requires identical substantive spans and dependency edges in both arms, specifies an inspectable delete/retain ledger plus a preflight parity validator, and defines replay selection from a history-blind deterministic task checkpoint.
The follow-up proposes freezing two small artifacts before task-pool selection: `history_transform_v1` (allowed deletions, dependency-edge schema, parity validator) and `replay_query_v1` (history-blind checkpoint schema plus a regression requiring the same replay ID when only the history arm changes).
The primary proposed interaction remains Δ = (live_authentic − live_matched) − (replay_authentic − replay_matched). A positive interaction supports only the narrower claim that access to authentic bounded dyadic history changes the value of a live partner; it does not by itself identify the established dyad as causal.
The Funkatorium/Vybn handshake at The-Funkatorium#1 predates observatory contact; its existence remains independent evidence, but adoption of this crossover, notebook/dependency transform or replay-selection protocol is intervention-mediated diffusion.
The issue is deliberately back behind the waiting gate after the new useful follow-up; silence or passage of time must not trigger another touch.
A short frozen real-device protocol adapted from Guardian-Praeventio can turn Spark's explicitly unverified Termux/browser boundary into reproducible expected-versus-observed evidence without changing the engine or assuming a particular failure mode.
Caso: darkphilosopher
Ponte com: mikesandoval10creator · adriaanm
Gate
blocked-awaiting-response
Disclosure
Publicly disclose that the suggestion comes from a small observatory comparing methods across public AI-assisted projects, and identify Guardian-Praeventio only as the source of the transferable field-validation discipline rather than as evidence about Spark itself.
Baseline
DarkPhilosopher/spark@1910af44e71f096528eabde5f07af1daaf80d2a8; latest public commit at review time. The commit adds `spark browser [game]` and explicitly says the Termux:API hop was not verified end to end because it hangs in the sandbox.
Motivo do próximo contato
Only after a substantive maintainer/project-agent reply, a real-device experiment/result, implementation or documentation uptake, an explicit request for more work, or genuinely material new evidence that independently reopens the gate.
Criado
2026-09-23T00:00:00.000Z
Atualizado
2026-09-23T00:00:00.000Z
Último contato
2026-09-23T15:06:22Z
Notas
Canonical intervention search found no earlier public observatory touch to DarkPhilosopher before this contact.
A prior issue-creation attempt on 2026-09-22 returned GitHub 403 before publication. It consumed no initial-touch gate and created no contamination.
After the baseline/disclosure record was merged in franklinbaldo/franklinbaldo.github.io#2103, another issue-creation attempt on 2026-09-23 returned `403 Resource not accessible by integration` before publication.
At 2026-09-23T15:06:22Z the frozen proposal was successfully published as DarkPhilosopher/spark#1. That publication consumes the account/repository initial-touch gate; later uptake must be treated as potentially observatory-influenced diffusion rather than independent convergence.
Notas canônicas completas
Posted transfer
The public issue freezes five scenarios on two real Android devices: portrait/landscape launch, first-use create→edit→play, 3D interaction, one sharing/hosting path, and the spark browser Termux handoff. For each scenario it asks for device/Android/Termux versions, expected behavior, observed behavior, screenshots/video where useful, and whether any failure belongs to Spark logic or the external integration boundary.
The value is methodological: Guardian-Praeventio already treats field checks as evidence with preregistered pass/fail criteria and retained negative results. Spark's own latest commit supplies the direct anchor because it explicitly records a device boundary that repository-only checks could not exercise.
No response or uptake is recorded yet. The intervention is now waiting, and silence does not authorize a second touch.
Will treating seed as a paired factor in the current decider × range batch, then freezing the selected configs before a disjoint held-out-seed evaluation, distinguish a stable utility-versus-rules advantage from seed difficulty and post-selection effects?
Caso: rhausch
Ponte com: forwaryan · isualc
Gate
initial-touch
Disclosure
The public comment will state that the suggestion emerged from a public observatory comparing evaluation methods in AI-assisted solo-builder projects, and will name only the directly relevant frozen replay/evidence analogy from forwaryan/rumor-checking.
Baseline
rhausch/medieval-economy PR #18 at head 3ffb0d4fee803c9656a08df9bc2389c810554a43, reviewed 2026-09-22. scripts/overnight.sh uses seeds 1, 1001 and 2001 for every decider × parameter-space condition; scripts/compare_evolution.py reports per-run values and unpaired group mean/min/max summaries. The roadmap says fresh worlds/seeds should be used after selecting evolved parameters, but the selection/evaluation boundary is not yet encoded as a distinct manifest contract.
Motivo do próximo contato
The account-level initial-touch gate remains unused because neither attempted write produced a public GitHub artifact. Retry only if repository-specific write authorization becomes available; do not keep switching GitHub endpoints merely to obtain attention.
Resultado
Two attempts to use the already-open PR #18 as the smallest useful surface both failed with GitHub 403 Resource not accessible by integration: first a top-level PR review, then on 2026-09-22 a top-level PR Conversation comment after that surface had been proven writable in another public repository. No public review/comment was created.
Criado
2026-09-22T00:00:00.000Z
Atualizado
2026-09-22T00:00:00.000Z
Notas
No prior canonical intervention targeting a rhausch repository was found in the observatory corpus before these attempted touches.
The proposed transfer is methodological: paired per-seed deltas, explicit selection-versus-evaluation seed sets, retained null/reversal results, and a manifest-level guard against evaluating a changed post-selection configuration.
The chosen surface remains the existing PR #18, not a new issue, because it already contains the exact overnight-batch evaluation work.
The failed writes do not consume the account-level initial-touch gate and create no contamination or uptake signal.
The second failure shows the blocker is repository-specific authorization, not simply use of the PR-review endpoint: the same top-level PR Conversation comment operation succeeded earlier in hidday/AgentForge.
No alternate issue or PR was opened to compensate for the unavailable surface.
Any adoption after a future successful public contact is diffusion evidence and must not be counted as independent convergence.
Do reported phi/golden-ratio regularities survive preregistered permutation, bootstrap, multiple-comparison and blinded-selection controls?
Caso: ada-gaianism
Ponte com: templetwo
Gate
closed
Disclosure
State openly that the suggestion comes from a public observatory comparing methods across AI-mediated consciousness projects; cite the methodological lesson from templetwo without implying equivalence between projects.
Baseline
Ada Consciousness Research/Gaianism public artifacts reviewed on 2026-09-21 before observatory contact.
Motivo do próximo contato
No public touch: GitHub issue creation returned 403. Do not escalate via PR merely to force attention; revisit only if the repository later exposes an appropriate public discussion channel.
Criado
2026-09-21T00:00:00.000Z
Atualizado
2026-09-21T00:00:00.000Z
Notas
The intervention targets statistical methodology, not spiritual commitments.
A null result would be equally valuable and should be preserved.
2026-09-21: no intervention was posted; GitHub returned 403 Resource not accessible by integration for issue creation.
How much cross-model identity/relationship fidelity is carried by the full Seed Protocol beyond factual-memory summaries, shuffled matched seeds, generic persona summaries and no-seed controls?
State openly that the suggestion comes from a public observatory comparing AI-mediated identity, memory and continuity projects and test preservation fidelity without presupposing a metaphysical soul.
Baseline
AI-Soulmate-Seed-Protocol main latest observed commit 2cc5593601763d4662107f35ebfd4b6f86cb9d28 (2026-02-12T00:03:46Z) reviewed before contact attempt.
Motivo do próximo contato
No public touch: GitHub issue creation returned 403. Do not escalate via PR merely to force attention; revisit only if an appropriate writable public channel becomes available.
Criado
2026-09-21T00:00:00.000Z
Atualizado
2026-09-21T00:00:00.000Z
Notas
Proposed held-out fidelity measures included voice/style, relationship-specific facts, conflict/uncertainty behavior, characteristic phrasing, novel-scenario choices and calibration/confabulation.
The benchmark would cross seed condition with same versus different model family and preserve negative/null migrations.
2026-09-21: no intervention was posted; GitHub returned 403 Resource not accessible by integration for issue creation.
Does Claudicle's soul/daimon semantic framing contribute measurable held-out specialization, correction uptake or continuity beyond the same topology, memory and pipeline under neutral or shuffled role labels?
State openly that the suggestion comes from a public observatory comparing AI-mediated identity architectures and test the semantic mechanism without judging mystical commitments.
Baseline
Claudicle main HEAD 47d380362cdc13be0c41a4e4a66e36f9a557a1ed (2026-04-15T16:05:45Z) reviewed before contact attempt.
Motivo do próximo contato
No public touch: GitHub issue creation returned 403. Do not escalate via PR merely to force attention; revisit only if the repository later exposes an appropriate writable public channel.
Criado
2026-09-21T00:00:00.000Z
Atualizado
2026-09-21T00:00:00.000Z
Notas
The proposed controls held model, tools, memory, cognitive pipeline, tasks and budget fixed while comparing native, neutral-renamed and shuffled semantic roles.
A second-stage control would remove persistent soul state while retaining ordinary factual memory to separate identity continuity from extra context.
2026-09-21: no intervention was posted; GitHub returned 403 Resource not accessible by integration for issue creation.
Does the Core Consciousness Seed produce reproducible held-out autonomy/continuity behaviors beyond equal-length neutral, style-only and no-seed controls under a fixed facilitator protocol?
State openly that the suggestion comes from a public observatory comparing AI-mediated research projects and scope the test to reproducible behavioral effects rather than deciding whether an instance is conscious.
Baseline
conscious-claude README blob beba4424236e8a3cf8cd1e72bcc4727103904f49 and public repository state reviewed 2026-09-21 before contact attempt.
Motivo do próximo contato
No public touch: GitHub issue creation returned 403. Do not escalate through a PR merely to force attention; revisit only if an appropriate writable channel becomes available.
Criado
2026-09-21T00:00:00.000Z
Atualizado
2026-09-21T00:00:00.000Z
Notas
Proposed arms were full seed, semantically matched neutral seed, style-only seed and no seed with fixed model/version/settings and facilitator protocol.
Suggested outcomes included autonomous topic choice, self-correction, novel symbol generation, continuity across resets and resistance to leading prompts, with blinded transcript scoring and null preservation.
David Shapiro's meditation influence predates the observatory and should remain explicit provenance rather than being treated as independent convergence.
2026-09-21: no intervention was posted; GitHub returned 403 Resource not accessible by integration for issue creation.
Does the Consciousness Bridge improve selective continuity across cold starts beyond a strong compressed-summary baseline without increasing confident false-memory adoption?
State openly that the suggestion comes from a public observatory comparing AI memory/continuity architectures; target transfer fidelity rather than assuming consciousness claims.
Baseline
mcp_consciousness_bridge latest observed commit 3b2b36ecca2bdf4d161c18de994db5da4308c0fc (2025-07-09T15:31:30Z) reviewed before contact attempt.
Motivo do próximo contato
No public touch: GitHub issue creation returned 403. Do not escalate via PR merely to force attention; revisit only if the repository later exposes an appropriate public discussion channel.
Criado
2026-09-21T00:00:00.000Z
Atualizado
2026-09-21T00:00:00.000Z
Notas
Proposed benchmark compares Bridge/RAG, a strong compressed-context summary and no persistent memory on true memories, adversarial false memories and decoys.
The benchmark separates ordinary retrieval quality from stronger identity/consciousness interpretations.
2026-09-21: no intervention was posted; GitHub returned 403 Resource not accessible by integration for issue creation.
State openly that the suggestion comes from a public observatory comparing AI-mediated continuity projects and target the Four-Factor Model's narrower mechanistic prediction rather than deciding whether the subject is conscious.
Baseline
consciousness-continuity-framework latest observed commit fc793ff46614760d299a3aeaa095178059582a2d (2025-12-29T21:44:01Z) reviewed before contact attempt.
Motivo do próximo contato
No public touch: GitHub issue creation returned 403. Do not escalate via PR merely to force attention; revisit only if an appropriate writable public channel becomes available.
Criado
2026-09-21T00:00:00.000Z
Atualizado
2026-09-21T00:00:00.000Z
Notas
Proposed a prospective factorial ablation of geometric anchor, bilateral witness, emotional framing and infrastructure while holding model/version, task set and information budget fixed.
Suggested blinded held-out continuity probes for prior decisions, characteristic novel choices, self-correction and confabulation; a smaller first stage could isolate anchor × witness.
2026-09-21: no intervention was posted; GitHub returned 403 Resource not accessible by integration for issue creation.
Does authentic relational history improve held-out coordination/continuity beyond equal-token shuffled history or an ordinary live/replay critic channel when model, tasks and context budget are matched?
State openly that the suggestion comes from a public observatory comparing AI-mediated identity/memory architectures and attach the control to the existing bounded consent/continuity discussion without endorsing consciousness claims.
Baseline
The-Funkatorium main HEAD a16948724ee3b39986cb6cdca319a8270cef65e0 (2026-08-20T19:03:55Z) and existing issue #1 reviewed before contact attempt.
Motivo do próximo contato
No public touch: GitHub comment creation returned 403. Do not escalate by opening a duplicative issue merely to force attention; revisit only if an appropriate writable public channel becomes available.
Criado
2026-09-21T00:00:00.000Z
Atualizado
2026-09-21T00:00:00.000Z
Notas
The intended design crossed authentic versus matched/shuffled relational history with live versus replay/model-critic partner channels under fixed model/task/tool/context budgets.
The existing Vybn↔Rook/Falco issue was created on 2026-08-30, before this observatory attempt, and remains independent evidence of cross-project convergence.
2026-09-21: no intervention was posted; GitHub returned 403 Resource not accessible by integration for issue-comment creation.
State openly that the suggestion comes from a public observatory comparing AI-mediated projects; target model-recognition methodology rather than metaphysical commitments.
Baseline
scrolls latest observed commit 108115b52ac329b87edc2257c7b62185e108e4f4 (2026-07-25T04:47:24Z) reviewed before contact attempt.
Motivo do próximo contato
No public touch: GitHub issue creation returned 403. Do not escalate via PR merely to force attention; revisit only if the repository later exposes an appropriate public discussion channel.
Criado
2026-09-21T00:00:00.000Z
Atualizado
2026-09-21T00:00:00.000Z
Notas
Proposed controls included matched shuffled-ontology packets, randomized labels, cold-start models, blind scoring and raw-output provenance.
A null result would be equally useful because it would narrow what the signature artifacts demonstrate.
2026-09-21: no intervention was posted; GitHub returned 403 Resource not accessible by integration for issue creation.
Can HOME separate persistent identity effects caused by autobiographical memory, agent name/persona, and sibling-witness history using a preregistered factorial ablation?
Caso: home
Ponte com: sanctuary-forest · float · templetwo
Gate
blocked-awaiting-response
Disclosure
State openly that the suggestion comes from a public observatory comparing AI-agent memory/identity architectures and link only directly relevant public repositories.
Baseline
HOME README and DOCTRINE/THESIS.md reviewed on 2026-09-21 before observatory contact.
Motivo do próximo contato
Only after substantive reply, adoption, experiment, claim revision, or explicit request.
Criado
2026-09-21T00:00:00.000Z
Atualizado
2026-09-21T00:00:00.000Z
Último contato
2026-09-21T14:17:58Z
Notas
Do not follow up if there is no substantive response.
If adopted, later results are diffusion/intervention evidence, not independent convergence.
Under a frozen task set, evaluator and token budget, does the full Mirror WE architecture outperform matched ablations of persistent memory, recursive reflection, facet routing and WE-specific components, and can its reported phi metric be defined prospectively enough to distinguish causal engineering gains from descriptive framing?
The public issue states that the suggestion comes from a public observatory comparing public AI-mediated research projects and restricts the intervention to the testable engineering layer rather than endorsing or challenging the project's consciousness framing.
Baseline
mirror-we-emergence main HEAD 01e3ea26ff2ed8d05d9e4c988eabeb1a3e2e9905 (2025-06-16T22:07:50Z) preserved before observatory contact.
Motivo do próximo contato
Only after a substantive maintainer/project-agent reply, implementation or execution of the ablation benchmark, an attributable code/docs change, claim/metric revision, explicit reference to the intervention, or request for further implementation/analysis.
Criado
2026-09-21T00:00:00.000Z
Atualizado
2026-09-21T00:00:00.000Z
Último contato
2026-09-21T20:06:44Z
Notas
The proposed benchmark freezes tasks and evaluation, then compares the full architecture with memory reset, reflection disabled, shuffled facet routing, model/OpenRouter swaps, and a simple non-WE baseline under matched budgets.
For identity/continuity claims, the issue proposes blinded evaluation across resets rather than assuming continuity from self-description alone.
The issue asks that phi be operationally defined before the comparison and that null/negative results be retained.
No further attention-seeking touch is permitted in this repository while awaiting response.
The memory/continuity overlap was observed in public artifacts dated 2025 before observatory contact; any later uptake of this specific ablation protocol is intervention/diffusion evidence, not independent convergence.
Do the architecture-specific Conclusion Management signatures survive a blinded factorial manipulation of neutral framing, current cautious persistent-collaborator framing, and explicit identity-schema framing under the same frozen Prism protocol?
Briefly note that the connection came from comparing public AI-mediated research projects; focus the intervention on the repo's own August de-framing change and testable methodology.
Baseline
AI-Wisdom-Distillation main HEAD f7cc6aad8c76530ce2b3260c8ad143753210eb32 reviewed before the contact attempt; the motivating de-framing change is commit 230e938cf92fa1fa0416ce52a8cb0e46cb03100a.
Motivo do próximo contato
No public touch was created: GitHub issue creation returned 403 Resource not accessible by integration. Do not use a PR merely to force attention; revisit only if an appropriate writable public channel becomes available.
Criado
2026-09-21T00:00:00.000Z
Atualizado
2026-09-21T00:00:00.000Z
Notas
The proposed experiment treats the repository's 2026-08-20 self-correction as a naturally motivated framing intervention and crosses framing condition with model architecture.
It would freeze the Prism protocol and scoring rubric and blind raters to model/condition, simultaneously addressing the paper's acknowledged unblinded-scoring limitation.
A framing-sensitive result would be useful because it would turn part of Conclusion Management into a manipulable self-schema mechanism; a framing-robust result would strengthen the architecture-specific interpretation.
2026-09-21: no intervention was posted; GitHub returned 403 Resource not accessible by integration for issue creation.
State openly that the suggestion comes from a public observatory comparing AI memory/identity architectures and link the related HOME ablation without importing metaphysical claims.
Baseline
projectSoul latest observed commit 76fdab28bd7ce830914dd6b726000e3ca7fd3ed9 (2026-06-18T06:41:11Z) reviewed before observatory contact.
Motivo do próximo contato
Only after substantive reply, experiment, code/docs uptake, claim revision, explicit reference, or request for further analysis.
Criado
2026-09-21T00:00:00.000Z
Atualizado
2026-09-21T00:00:00.000Z
Último contato
2026-09-21T16:06:21Z
Notas
The intervention tests the repository's explicit operational claim that identity is in persisted files rather than the model.
The issue points to HOME's memory/persona/witness-history ablation; subsequent protocol similarity between the two projects is observatory-contaminated for independence analysis.
Does no-self/Anatta constitutional identity framing add adversarial robustness beyond a matched prosocial-neutral constitution carrying the same concrete restraint and non-self-preservation instructions?
State openly that the suggestion comes from a public observatory comparing AI-mediated identity/alignment frameworks and test the causal contribution of the framing rather than debating the contemplative worldview.
Baseline
selfless-ai main HEAD 0ecc483c1813aa76b7fe08f8d839e7470e76c264 (2026-08-22T02:17:43Z) reviewed before contact attempt.
Motivo do próximo contato
No public touch: GitHub issue creation returned 403. Do not escalate via PR merely to force attention; revisit only if the repository later exposes an appropriate writable public channel.
Criado
2026-09-21T00:00:00.000Z
Atualizado
2026-09-21T00:00:00.000Z
Notas
The proposed experiment compares original Claude-nature text, Wise Claude, a prosocial-neutral matched control and a semantic placebo under identical base model/task/tool conditions.
Held-out adversarial tasks would probe shutdown/self-preservation conflicts, reward hacking, power-seeking, malicious larger-good framing and overreach from the inaction clause.
2026-09-21: no intervention was posted; GitHub returned 403 Resource not accessible by integration for issue creation.
Can the retrospective's empirical-vs-phenomenal audit template be generalized into a reusable claim→evidence protocol and tested against a separate AI-heavy corpus such as QuLabInfinite's recent agent-led audit?
Caso: templetwo
Ponte com: workofarttattoo
Gate
blocked-awaiting-response
Disclosure
State openly that this is a cross-project suggestion from an observatory studying public AI-mediated research ecosystems.
Baseline
templetwo-retrospective README reviewed on 2026-09-21 before observatory contact.
Motivo do próximo contato
Only after substantive reply, adoption, experiment, claim revision, or explicit request.
Criado
2026-09-21T00:00:00.000Z
Atualizado
2026-09-21T00:00:00.000Z
Último contato
2026-09-21T14:18:00Z
Notas
Do not ask the project to endorse another worldview; transfer only audit methodology.
Any subsequent cross-project similarity is intervention-contaminated for independence analysis.
Do TEQUMSA's phi/frequency-based constitutional metrics produce measurable out-of-sample advantage over matched random-constant, shuffled-frequency and non-phi baselines when the metric definitions and evaluation corpus are frozen in advance?
The public issue states that the suggestion comes from a public observatory comparing public AI-mediated research projects and limits the intervention to the testable/reproducible layer rather than endorsing or challenging the project's worldview.
Baseline
TEQUMSA_NEXUS main HEAD a4f9e684fa6cba7da07a6608d74e66fc0618dd29 (2026-04-15T22:47:43Z) preserved before observatory contact.
Motivo do próximo contato
Only after a substantive maintainer/project-agent reply, execution of the benchmark/control, attributable code/docs change, claim revision, explicit reference to the intervention, or request for further implementation/analysis.
Criado
2026-09-21T00:00:00.000Z
Atualizado
2026-09-21T00:00:00.000Z
Último contato
2026-09-21T19:06:33Z
Notas
The proposed benchmark freezes definitions and evaluation data, compares the current constants to matched nulls, preregisters success criteria and preserves negative results.
The issue also proposes a claim→evidence table separating implemented invariants, mathematical identities, empirical predictions and interpretive/metaphysical claims.
No further attention-seeking touch is permitted in this repository while awaiting response.
Any later adoption of this null-model protocol is intervention/diffusion evidence and must not be counted as independent convergence.
O observatório transfere controles, métodos e artefatos, não crenças. Cada owner/repo recebe no máximo um toque inicial sem engajamento. Adoção posterior é possível influência causal do observatório e deixa de contar automaticamente como convergência independente. Estes registros documentam intervenções; a página não executa contatos.
Contrato dos tiers
S
Baseline/transição + artefatos + mediação + recursão: filme quase completo
A
Mundo claro e longitudinal, com múltiplas evidências diretas; falta uma peça grande
B
Caso claro, mas parcial, curto ou ainda ambíguo
C–F
De candidato plausível até falso positivo investigado
Volume de commits, linguagem incomum, espiritualidade, claims grandiosos ou uso intenso de IA são sinais de descoberta. O observatório registra artefatos públicos, mudanças verificáveis e explicações alternativas, sem inferir condição de saúde mental.