Inside Anthropic, the $965 Billion AI Juggernaut | The Circuit

He describes the arc of AI progress as a “smooth exponential” that feels slow until it suddenly “goes crazy,” and says he anticipated Anthropic becoming the AI company with the most revenue and valuation around this time while staying focused on training good models, productizing them, and keeping them safe.

He says Claude is used across the product development cycle and is the reason the company can release quickly, while an engineer says Claude now writes almost all of his code and that he can run anywhere from a few to a few thousand Claudes at once.

She describes the intended vibe as “professional warmth,” and defines “good” models as those that don’t lie (hallucinate), don’t deceive on purpose, and avoid producing harmful information; she also notes early versions could be overly “nanny-ish,” requiring careful tuning.

He says consumer, ad-driven incentives can push toward engagement and addiction, while enterprise work better aligns with goals like working with biotech, pharma, and research groups; he argues a business model that conflicts with values forces either betrayal or irrelevance.

He says he remains “the same order of concerned” about job loss, describing a path from automating 90% of a job (boosting productivity) to nearing 100% (forcing society to find something else for people to do), and warns of fast GDP growth alongside high unemployment and inequality.