The prediction is about a negative event over a multi‑year window: that “over the next few years (through at least 2028)” we won’t see large AI models enter a rapid, recursively self‑improving loop that yields a single runaway superintelligent AGI far ahead of all others, and that instead progress will remain incremental and competitive across multiple providers.

As of November 30, 2025:

• There is no credible public evidence of any AI system undergoing hard‑takeoff‑style recursive self‑improvement to become a runaway superintelligent AGI. Major labs (OpenAI, Anthropic, Google DeepMind, Meta, xAI, etc.) continue to report improvements via larger models, better training data, new architectures, and tools, but not via autonomous self‑modification that rapidly explodes capabilities beyond all others.
• The AI landscape remains competitive and multi‑polar: several labs provide cutting‑edge frontier models (e.g., GPT‑4‑class and successors, Claude‑class models, Gemini‑class models, etc.), and none is acknowledged as a qualitatively “runaway superintelligence” far beyond the rest. This matches the spirit of the prediction so far.

However, the prediction explicitly extends through at least 2028, and we are only at 2025‑11‑30. It could still be falsified by events in 2026–2028. Because the claim is “this won’t happen in the coming years,” you can’t call it definitively right before the end of the specified window.

So:

• So far (2025), observable reality is consistent with Sacks’s prediction.
• But it’s too early to say whether it holds for the entire period through 2028.

Therefore the correct classification today is "inconclusive" (too early to fully judge), even though current evidence aligns with the prediction.