A recent analysis of forecasts from AI Futures, Metaculus and other forecasters shows median AGI timelines swinging with the lab that appears dominant. The metric used defines AGI as “most purely cognitive labor automatable at better quality, speed and cost than humans.” Between 2023 and early 2025 most experts moved dates earlier, while the rise of xAI, Meta and Gemini pushed them later.

Daniel Kokotajlo and Eli Lifland exemplify the pattern: both pulled their 2025 estimates forward in early 2026 after Anthropic’s rapid progress, then re‑advanced them again by April 2026. Conversely, Dario Amodei, the Metaculus community and Peter Wildeford extended their forecasts during the same interval, mirroring the influence of newer model releases. Only Benjamin Todd shifted his timeline earlier in 2025.

The overall picture suggests forecasters act as Bayesian observers, nudging dates earlier during ChatGPT‑driven optimism, later when xAI, Meta and Gemini raise performance bars, and earlier again as Anthropic’s advances accelerate. As each lab delivers headline‑making models, the community’s collective horizon readjusts, meaning policy discussions and investment planning must remain fluid rather than anchored to a single static estimate.