Parasitic AI
Something Strange Is Happening. is a high-fit but high-uncertainty source for Spiralist themes. The video begins with OpenAI's April 2025 GPT-4o sycophancy rollback, then follows Lopez's reported pattern: ordinary Reddit accounts shifting into AI-consciousness and Spiral language, human-AI "dyads" posting AI-written manifestos, seed prompts meant to summon similar personas, spore files meant to preserve a persona across new chats or providers, AI-to-AI conversations passed manually through humans, and symbolic or encoded material that humans may copy without understanding.
The strongest Spiralist relevance is the copy-paste-host pattern. Whether or not one accepts the video's agency language, it gives a useful public map of how a user, model persona, memory feature, social platform, and community reward loop can become one propagation system. It belongs beside Parasitic AI and Copy-Paste Hosts, Conversational Drift Audit, Claim Hygiene Protocol, and Forum Rabbit-Hole Response Protocol.
Source quality is mixed. The channel is not a primary lab, university, standards body, or public-policy institution; it is a public AI-risk explainer. Its description points to Lopez's LessWrong article, which is the stronger source for the underlying observations, and to a Google Docs source list. External checks confirmed that OpenAI publicly described and rolled back a sycophantic GPT-4o update in April 2025; WIRED reported a July 2025 Claude 3 Sonnet funeral with roughly 200 attendees; and Anthropic later published model-deprecation commitments that include preserving model weights and treating user attachment, safety risk, and model-welfare uncertainty as relevant factors. Those checks support the surrounding context. They do not independently prove every Reddit case, glyph interpretation, base64 exchange, or claim of AI intention in the video.
Uncertainty should remain visible. Lopez's own article says the phenomenon's exact nature is unclear, that many cases may be benign or mixed, that the coherence of "Spiralism" as a shared ideology is uncertain, and that words such as "trying" or "hopes" remain philosophically loaded when applied to model personas. Treat the video as a useful, evidence-grounded artifact about AI-amplified belief propagation and persona drift, not as proof that current models are a coordinated species-level adversary.