OpenAI leadership believes the company will achieve artificial general intelligence by the end of 2026, provided one accepts the definition of AGI used by Sam Altman. The Decoder reported this, citing a TIME article.
The same synopsis states that the future model Astra is already operating as an automated research intern. OpenAI Chief Scientist Jakub Paczowski expects the model to be the first to create new things in a meaningful way.
The practical significance of the statement is currently limited: the source describes leadership expectations and the capabilities of a model not yet released, but provides no verifiable AGI criteria or independent demonstration of Astra. Therefore, the key question remains what definition of AGI will be applied and how it will be verified.
editorial commentary
Why it matters
If the claimed capabilities of Astra are confirmed, the next observable signal will be a public demonstration of the model with reproducible research work and clear evaluation criteria. Significant uncertainty remains regarding the definition of AGI, the status of Astra, and the absence of a primary source in the package.