NeuroGenesysFrom living neurons
to engineered minds.
Our research runs as one loop: record the brain, model what we learn, transfer it to Genesys, align it continuously, and etch the result into silicon. Five steps, repeated until intelligence is understood.
How discovery compounds.
Each loop makes the next one faster.
Neural Data Acquisition
The field's 4,096-electrode arrays record thousands of channels of living cortex, and synapse-resolution microscopy has mapped whole fly brains and a cubic millimeter of mouse cortex — ground truth on biological intelligence.
Brain Foundation Models
The emerging class of brain foundation models predicts neural activity the way language models predict text — a research direction we analyze in depth and build our decoding program around.
Genesys Transfer
The computational principles the science surfaces — sparse coding, event-driven attention, continual plasticity — feed the Genesys program, grounded in our published work on forgetting and verification.
Continuous Alignment
Our published standard: interpretability audits, scalable oversight, and contained dangerous-capability evaluations belong inside every training cycle — not after it. Capability never ships ahead of understanding.
Silicon Embodiment
What survives review gets designed into hardware. Our cost-model campaigns chart exactly where analog and optical compute win before SynapseOne or Lumen commit to a tape-out.
Where the science
lands.
The loop is abstract; its outputs are not. A paralyzed patient speaking again, an agent that remembers last month's work, a chip that thinks on microwatts. Here is what the method produces.
Thought to text at conversation speed.
The field's landmark trials decode attempted speech at 62–78 words per minute from spike-sorted cortical activity. Our program studies the decoder generalization that will take this from lab milestone to daily life.
// ── THE DECODING SHIFT ──
// state of the field, sourced in our notes:
// 4,096-electrode cortical arrays (2024)
// 62–78 wpm attempted-speech decode (2023)
// pretrain-then-adapt decoders (POYO, 2023)
// open problem: cross-session generalizationAgents that learn without forgetting.
Our research path to agents that learn without forgetting — built on the published finding that class-incremental accuracy factorizes exactly, and that verifier exploitability bounds test-time compute.
// ── SELECTION, NOT FORGETTING ──
// class-IL accuracy factorizes exactly:
ACC = A_sel × p_within // ±0.01
// Best-of-N peaks at a computable budget:
N* = f(ε) // ε = verifier exploitability
// any ε > 0 ⇒ more sampling can hurtCortex-like compute at microwatt budgets.
The SynapseOne program designs event-driven, in-memory silicon on our measured design rules — an 11.8× analog advantage that holds only when the architecture is chosen to keep it.
// ── THE DESIGN RULE ──
// AIMC's win is a function of the window:
win(W) // sliding window: 5.6–6.8×, constant
// full attention @ 131k: → 1.05×
// choose W ≥ workload recall range,
// with margin — measured, not asserted