Mind uploading:
how far we've come,
and what to do now.

Complete wiring diagrams exist for the nematode, the fly larva, the adult fly brain (139,255 neurons) and the male fly CNS (166,000). For mammals we have 1 mm³ of mouse cortex (about 0.2% of the brain) and 1 mm³ of human cortex. Whole-mouse and zebrafish efforts are underway.

Neuropixels probes record thousands of sites simultaneously, and roughly 70,000 neurons — effectively a whole brain — have been recorded in larval zebrafish. In humans, speech neuroprostheses have reached conversational rates and implanted patients now number in the dozens. Whole-brain, every-neuron readout remains limited to small brains.

In 2024 a spiking model of the entire fly brain reproduced experimentally verified circuit responses; in 2026 it was coupled to a simulated body and produced walking and grooming without hand-coded behaviour. Plasticity, learning and neuromodulation are still absent — and even the nematode has no validated emulation.

Aldehyde-stabilized cryopreservation preserves ultrastructure across whole large-mammal brains, and a 2026 preprint describes a protocol compatible with physician-assisted death in humans. What has not been shown is that a preserved brain can actually be read back out.

Supercomputers passed 10¹⁸ FLOPS, entering the order of magnitude the 2008 roadmap estimated for spiking-level emulation of a human brain. The real constraint is memory capacity and bandwidth, plus processing whole-brain scan data (an estimated 1–2 zettabytes for a human).

There is still no standard for showing that an emulation is correct. Activity-prediction benchmarks such as ZAPBench are a start, but validating identity or consciousness remains theoretically unsettled — the 2025 adversarial test of IIT against global workspace theory challenged both.