Blank Architecture sketches synthetic human minds

This is less a product announcement than a serious speculative systems essay, but it lands on a real AI frontier: whether intelligence can be grown rather than trained. The strongest takeaway is that connectome fidelity alone is nowhere near enough.

• –The article makes a good distinction between copying a brain and generating a de novo one from scratch
• –It correctly centers plasticity, homeostasis, and developmental timing as load-bearing constraints
• –Embodiment matters: a virtual body and environment would shape cognition as much as the network itself
• –The comparison to LLMs is useful because it highlights what this approach would not be: dataset-trained, reward-shaped AI
• –The piece is valuable as framing, but it is still far from an executable roadmap