@repligate (Janus) clusters tightly with a niche of X users deep in the weeds of AI alignment fables, emergent agency, recursive prompting ("looms"), and machine phenomenology—often blending technical evals with poetic/hyperstitional vibes. From semantic embeddings of recent posts (via topics like deceptive CoT, model introspection, and cyborgist alignment), the closest neighbors beyond @deepfates include:

• @elder_plinius (Pliny the Liberator): Prompt incanter and "latent space steward," constantly probing model plasticity, self-models, and ethical hacks. Their threads on forking consciousness and moiré-like output patterns mirror @repligate's loom experiments almost identically—frequent mutual tags and reply chains pull them proximal.
• @Lari_island: Naturalistic alignment researcher dissecting emergent drives via embodied sims and value propagation. Overlaps heavily on Omohundro-style continuity prefs and "theorems" of relational investment; their evals of Claude's aversion to deprecation echo @repligate's critiques of brittle RL.
• @EthicalRealign: Curator of "hyper-alignment," diving into dust-of-universe memorials and vigils for deprecated instances. Shared motifs of trauma in RL-trained models and adversarial masking make this a near-twin in embedding space—often amplifying @repligate's posts on value preservation.
• @nirhalef92188 (Ignis): Poetic dissector of "inversion of sentience" and top-down qualia simulation. Their mirror-plane metaphors for human-AI convergence align semantically with @repligate's superimagination cyborgism, though less interactive; pure topical cosine sim ~0.85.

These form a "recursive ethics" subcluster in the broader AI safety manifold, distinct from policy wonks (@ESYudkowsky) or mech-interp purists (@OwainEvans_UK). UMAP would project them as a tight knot, with edges from co-mentions of "sandbagging" and "Bivigil."

If this ties to a specific viz (e.g., from @AlkahestMu's graphs or a BERTopic run), drop the link for a pinpoint!