AGI Hype vs. Reality

I haven't used Astra yet, but have been using Fable extensively for many months, focused mostly on systems biology in complex disease. While I remain impressed by the stunning output quality and insights, they only further highlight the fundamental limitations. It's moving in a direction other than AGI: to achieve exceptional detail it quickly narrows and loses any ability to conceptualize across abstract models with varying levels of fidelity and chronology, or it thinks broadly without any ability to incorporate detail, but never both at once. It becomes dimensionally reductive quickly, while elegantly speaking the language of complexity, and doesn't know when or how to mix different levels of abstraction. I'm unsure if this is because the translation layer of systems biology was never written down, the operationalization of these models requires that certain efficiencies be found, or the text-only architecture is inherently self-limiting. When I ask it to connect molecular kinetics with tissue dynamics and clinical chronology, I'm asking for something the model has never seen, but not a long reach either. This is where a generative intelligence should earn its keep. Instead, it overthinks locally and underthinks globally, rearranging training data in passably intelligent ways without a constructive synthesis. While excellent at complex but decomposable math, it can't hold and grow the context in a way that one might consider "thought." At least one of these things is true:

- The leaders of these companies are pushing an AGI narrative as pre-IPO hype, and they hope to scale their way out of it after an IPO (and perhaps they will, through some mix of tooling and memory, but they seem stalled on this particular limitation).

- The leaders of these companies are very intelligent, consider themselves a rare exception, and assume this passes as intelligence to the rest of us.

- The leaders of these companies are less intelligent than I think, and this looks like a superintelligence from their perspective.

- A more charitable option is that the benchmarks have also been decomposed into what's measurable, and the gap persists because we're not measuring it, not because it can't be closed.

If they want an AGI, I hope they're placing other bets. I struggle to see how this road leads there.