The pursuit of mind uploading has reached a pivotal moment. The State of Brain Emulation Report 2025, authored by Niccolò Zanichelli and colleagues, represents the first major update to the field’s technical roadmap since Sandberg and Bostrom’s foundational 2008 document. After 17 years of progress, where do we actually stand?

The Roadmap vs. The Traffic

The report organizes its assessment around three pillars: Neural Dynamics (recording function), Connectomics (mapping structure), and Emulation (computational implementation). The findings reveal a landscape of exponential success paired with stubborn obstacles.

The Connectomics Triumph: Our ability to map the static structure of the brain has exceeded expectations. We have achieved accurate connectomes with verified synaptic connections for complex organisms like Drosophila melanogaster. The resolution of electron microscopy and the speed of AI-assisted image segmentation have made mapping the wiring diagram a solvable engineering problem. Recent innovations like SmartEM further accelerate this progress by using machine learning to guide microscopes, achieving 7x speedup while democratizing access to brain mapping technology.

The Dynamics Bottleneck: Conversely, the report identifies a critical limitation in our ability to record dynamics. Despite progress, no single-neuron whole-brain imaging capability exists that can capture the spiking activity of every neuron simultaneously with sufficient temporal resolution. We can map the roads, but we cannot yet see the traffic at scale.

Redefining Minimal Emulation

Perhaps most significant is the report’s redefinition of what a “Minimal Brain Emulation” actually requires. The authors move the goalposts considerably. They argue that a simple point-neuron model, which simulates neurons as basic logic gates, is insufficient for functional emulation.

True emulation likely requires simulating:

• Neuronal compartments (dendritic computation)
• Neuromodulation systems (volume transmission of dopamine, serotonin, and other chemicals)
• Fast subcellular processes that influence neural signaling

This suggests the computational load for a single brain is orders of magnitude higher than estimated in 2008.

The Strategic Pivot to NeuroAI

The most notable strategic shift proposed in the report is the alignment of Whole Brain Emulation research with AI Safety. The authors argue that understanding the biological brain is the safest path to building aligned Artificial General Intelligence. By studying the only known working model of general intelligence, us, we can derive principles for safe AI architecture.

This represents a maturation of the field. Rather than rushing toward immediate uploading capabilities, the scientific community is prioritizing understanding first.

Our Perspective

The 2025 report is not a setback. It is a recalibration based on empirical data. The fact that connectomics has progressed faster than expected demonstrates that the engineering challenges are solvable. The dynamics problem is complex, but complexity has never stopped scientific progress. It only determines the timeline.

For those of us working toward digital preservation of consciousness, this report provides clarity. The path forward requires deeper integration of neuroscience with computational modeling. Recent developments in Virtual Brain Twins demonstrate this integration in clinical practice, where patient-specific brain models are already being used to predict treatment responses. The destination remains the same. Only our understanding of the journey has improved.

---

Source: Zanichelli, N., et al. (2025). State of Brain Emulation Report 2025. arXiv:2510.15745 [q-bio.NC]. https://arxiv.org/abs/2510.15745