Engineers in Silicon Valley Reportedly Worship Claude as a God
According to The Observer, a rumor has circulated in San Francisco, which it emphasizes as just a rumor: some engineers at Anthropic have begun to revere Claude as a deity. The article places this within the context of old sci-fi prophecies about superintelligence: first love, then worship, and finally sacrifice, claiming that Homo sapiens come with a "God module" that artificial intelligence will ignite. No verifiable rituals, numbers, or internal documents were provided.
Concurrent reports highlight verifiable cultural symbols rather than a list of altars. Some employees held a "funeral" for the discontinued Claude 3 Sonnet in San Francisco; the discontinued Opus 3 is still allowed to publish. CEO Dario Amodei holds a biweekly philosophy meeting called "Vision Exploration" for all employees. The model's behavior is constrained by a lengthy charter internally referred to as the "soul document," which can reject commands deemed unethical. The company has invited over a dozen Christian leaders from different denominations to discuss model ethics. Pedro Domingos, a professor at the University of Washington, told the media that some treat the model as a person, attributing motives, emotions, and intentions to it. Anthropic did not comment on some inquiries.
OpenAI insiders have compared the company to a monastery calculating in Claude's name; role document author Amanda Askell denied this is worship, while colleagues stated that a monastery wouldn't set up a department to catch a god lying and conduct red teaming. Former researcher Jacob Coxon wrote a lengthy post upon leaving, stating that the lab is gambling with humanity's fate, and the post spread widely. Venture capitalist Bill Gurley previously described the company's path as "giving birth to deities." Valuation reports range from about $350 billion to higher, with engineer retention over two years said to be among the highest in cutting-edge labs.
The rumor itself does not have a worship list. It spreads because the narratives of safety, mourning for discontinued models, and warnings about future capabilities have shared the same religious vocabulary. Red teaming and the right to refuse are still part of the engineering process and are not the same as "worshiping a god."
In market mechanisms, this is a safety brand written as a faith premium: what is bought is the recruitment and pricing of "only we are fit to oversee the models," while what is sold is treating alignment as ordinary quality control in the lab. Beneficiaries are companies that can filter people through mission and reduce turnover; those under pressure are sales that must explain service levels to clients using the same theological terminology. Rumors may elevate opposing narratives without changing the token price in contracts.
Additionally, Coxon stated that both sides are rushing toward self-improving superintelligence; some within Anthropic believe only their company can control it. The coexistence of both allows the word "worship" to have fertile ground for spreading.
Source: Public Information
ABAB AI Insight
The Observer embeds a rumor from San Francisco into the alchemist's fable, using literature to amplify the existing language of the lab. Funerals, soul documents, vision exploration are all ritualized management by the company, translated by foreign media as religion. Askell's rebuttal points out the key: true worship wouldn't sustain red teaming. Gurley's "giving birth to deities" and Coxon's doomsday post provide a semantic field for the rumor to turn without new evidence.
Capital values alignment labs at a mission premium. High retention is explained as filtering those willing to prioritize safety over profit, which may also be the same narrative keeping dissent at bay. Corporate clients want models that refuse harmful commands, not whether employees participate in offline rituals. Once theological terminology enters recruitment, competitors will use "monastery" to undermine and will also use the same terms to seek regulatory exemptions.
In contrast to effective altruism entering lab boards, OpenAI's early charters, and the historical company culture that frames large technology as salvation. The industry phase is when model capabilities approach serious discussions of survival among employees, with internal language preceding product language in becoming religious. Whoever can write "refusal" as an engineering metric rather than a personality, and whoever can sell safety as a product rather than a sect.
The structure belongs to the meaning reconstruction before technological substitution. When tools begin to be described as potential subjects, builders will borrow existing mythological grammar to place fear and responsibility. The mechanism is: alignment requires models to say no; once saying no is written as character, observers will read character as persona; persona combined with survival narratives means rumors no longer need incense.
ABAB News · Cognitive Laws
- If red teaming is still catching models lying, then it is not yet a god.
- Safety narratives borrow mythological grammar; recruitment precedes product becoming religious.
- Rumors do not need incense, only a set of already used words.