Path A · Investors

Everyone is building the body. We're building the brain.

A spiking neural network that learns continuously, with mechanisms designed to protect old skills while learning new ones, and is designed to run on a fraction of the power of today's AI. The brain is real and training right now. The body is what we build next, with you.

1.16M neurons 814M synapses Patent pending

Market signals

$1.4T
Robotics market by 2030
Goldman Sachs
$200M+
Brain-inspired AI VC in 2025
3× from 2024
1.16M
Neurons training live
Today, not pitched
Few
SNN players building robot brains
Most sell chips, not brains

Why now

The market is moving beyond LLMs.
The biggest names in AI agree.

But their solutions still have the same old problems.

  1. March 2026

    Yann LeCun's AMI Labs raises $1.03B for "world models"

    The Turing Award winner left Meta to build JEPA, world models that go beyond LLMs. Nvidia, Toyota, Samsung, Jeff Bezos invested. The thesis: LLMs are not enough for robots.

  2. June 2026

    Jeff Bezos's Project Prometheus raises $12B at $41B valuation

    Bezos + Vik Bajaj as co-CEOs. JPMorgan, Goldman Sachs, BlackRock in. Founding advisors include Ashish Vaswani and Jakob Uszkoreit, two of the original "Attention Is All You Need" authors. Pitched as "physical AI" for industrial design and manufacturing. Wall-Street-scale capital is now pouring into physical AI.

AMI and Prometheus both still train on GPU clusters, still deploy frozen models, still can't learn on-device, and still have no answer for catastrophic forgetting. OSCEN is designed to answer all four.

AMI (JEPA)
GPU pretraining
No continual learning
Prometheus
VLA transformer
Frozen, GPU-scale
Figure / π
Transformer VLA
Static, 50-200W
Google
Gemini VLA
Retrain to update
OSCEN
Spiking Neural Net
Continual, <5W target

The energy problem

Robots need intelligence that runs on batteries,
not data centers.

Current AI is unsustainable for edge deployment.

GPT-4o Epoch AI, 2025
~1,080 joules per query
Gemini Google official
~864 joules per prompt
GPU Robot Brain Industry standard
50-200 joules per second
OSCEN on neuromorphic silicon Estimate from published SNN benchmarks, not measured
<5 joules per second (projected)

Event-driven compute only runs when neurons fire, not idle GPU cycles.

The architecture is ready; the compute layer is catching up.

1 watt = 1 joule per second. LLM figures are energy per interaction (~0.3 Wh and ~0.24 Wh converted to joules); robot brains draw power continuously.

Patent protection

The architecture is protected.
The IP moat is real.

4 Patents Pending

Full-Stack IP Portfolio

Filed Feb & Oct 2026 · USPTO

US 63/986,737

Filed Feb 20, 2026

Core Neuromorphic Cognitive Architecture

Integrated 6-mechanism biological learning, 5 developmental critical periods, multi-compartment dendrites, cross-modal binding, and hybrid SNN-LLM oracle.

US 64/184,488

Filed Oct 26, 2026

Body-Agnostic Robotics & Safety Kernel

Self-describing robot interface, per-body hardware safety arbitration, cross-body skill transfer, demonstration teaching, governed self-modification, and physical reflexes.

US 64/184,586

Filed Oct 26, 2026

Reward-Credit, Vector-Credit & Verification

Credit-gated consolidation, routed dopamine, typed aversive learning, extended 6-channel neuromodulation, vector-credit motor control, and causal verification batteries.

US 64/184,614

Filed Oct 26, 2026

Neuromorphic Silicon, Photonic & 3D Hardware

Developmental synaptic compression, 3-tier thalamic capsule paging, quantum-on-chip stochastic plasticity, split PQC + PUF security, photonic WDM, and 3D bare-die packaging.

Foundation Architecture Claims (US 63/986,737)

Claim 1

Integrated Multi-Mechanism Learning

6 learning rules operating simultaneously across the network

Claim 2

Developmental Critical Periods

5 phases with experience-dependent adolescent entry: pruning, myelination, identity

Claim 3

Hybrid SNN-LLM Architecture

Brain learns when to query external knowledge through emergent STDP, not hardcoded rules

Claim 4

Cross-Modal Binding

Arbitrary sensor modalities bind through spike-timing coincidence with multiplicative instinctual gain

Claim 5

Multi-Compartment Dendrites

4 compartments per neuron with independent dynamics and supralinear dendritic spikes

Claim 6

Neuromodulatory Continual Learning

Eligibility traces + homeostatic scaling + myelination enable learning without forgetting

4
U.S. provisional patents filed
258
Prior art patents searched
0
Overlapping claims found

What's built

This is not a pitch deck.
The brain is running.

1,156,800
Spiking neurons
+

15 brain regions. 65 synapse groups. Sensory, motor, association, prediction, working memory, all connected and learning.

814M
Synapses
+

Sparse CSR matrices. Eligibility traces on every plastic group. BCM, neuromodulation, and synaptic scaling active system-wide.

~10 years
Research and design
+

Nearly a decade of neuroscience research and architectural design. 7 months of focused implementation, bootstrapped with considerable personal investment.

24/7
Continuous training
+

Enterprise-grade dedicated server. Processing a multimodal video and audio curriculum around the clock. Currently in the adolescent developmental phase.

6
Learning rules
+

All running together, continuously. STDP, eligibility traces, BCM, neuromodulation, homeostatic scaling, R-STDP.

5
Developmental phases
+

Infant → Toddler → Juvenile → Adolescent → Mature. Phase transitions are experience-dependent, not hardcoded.

Training live · Adolescent phase

Commitment

Built with
conviction.

3,500+ hours
Dedicated R&D
~10 years
Research & development
Live
Brain training daily
CPU today
Low-power next

Deep domain knowledge across computational neuroscience, distributed systems, sparse linear algebra, and real-time sensor fusion. The architecture runs on CPUs today and the brain is actively training. Funding continues the build, opens partnerships with low-power compute providers, and gets us to purpose-built edge deployment.

Use of funds

What funding
enables.

  1. 01

    Low-power edge deployment

    +

    Move from CPU simulation to purpose-built edge compute. Target: up to 100x lower energy, based on published neuromorphic results.

  2. 02

    Robotics integration

    +

    Physical robot body. Camera, servos, touch sensors. Close the full sensorimotor loop with real-world feedback.

  3. 03

    Team expansion

    +

    Computational neuroscientist, robotics engineer, hardware specialist. Three hires take the build from a four-person core to a full build team.

  4. 04

    Patent portfolio

    +

    Non-provisional filing by Feb 2027. 20 continuation patents covering hardware mapping, fleet learning, safety architecture.

  5. 05

    Benchmarks & papers

    +

    Formal evaluation on standard robotics tasks. Published results for academic credibility and partnership conversations.

Team

The people
building OSCEN.

Deep expertise on the architecture, paired with the operational footing to scale it.

Rio Bold

Rio Bold

Cofounder & Architect

Marine Corps veteran. 13 years of professional software engineering across multiple Fortune 500 companies, building AI integrations, enterprise business processes, and full-stack systems at scale. Combined with nearly a decade of independent research in computational neuroscience, distributed systems, and real-time sensor fusion.

5 years of planning and architectural design, then 7 months of focused build to bring the complete OSCEN architecture from concept to 1M-neuron deployment.

Domain expertise: Spiking neural network theory, STDP/BCM learning rules, sparse matrix computation (CSR/COO), real-time event-driven systems, neuromorphic hardware constraints (Loihi, SpiNNaker), robotic motor control, and sensor fusion pipelines.

Every architectural decision. Every learning mechanism. Every line of code. Sensory gateway, brain simulation, motor output, cognitive action channel, speech motor, developmental phases, persistence layer, monitoring pipeline, and 5,000+ automated tests.

Adam Hamersky

Adam Hamersky

Cofounder & Operations

Marine Corps veteran with a record of leading people and running fleets under pressure. Served as Chief Motor Pool Dispatcher for the 7th Marine Regiment, coordinating personnel, vehicles, and mission readiness against tight operational windows.

Brings that same discipline and mission focus to building the brain that will move the next generation of robots, turning architectural ambition into a fundable, hireable, deployable company.

Operating focus: fleet logistics, program management, partnership development, and the operational footing required to take OSCEN from research artifact to commercial deployment.

Every operational decision. Every hire. Every milestone. The mission discipline that keeps the build on track.

Robert Robancho

Robert Robancho

Partner, Intelligence & Threat

Marine Corps veteran with a record of evaluating and neutralizing asymmetric threats in high-stakes environments. Built foundational expertise in tactical military intelligence, reading adversary intent under time pressure and translating ambiguous signals into defensive posture.

Transitioned to the consulting sector as an analyst at Deloitte, where he led complex operational logistics and data tracking for United States Space Force contracts. Recognized with the 2026 Charlie Allen Achievement Edwin Land Industry Award for outsized impact on the national security mission.

Operating focus: intelligence frameworks, threat modeling, OPSEC discipline, security clearance management, and hardening emerging physical AI platforms against modern vulnerabilities. Bridges ground-level tactical realities with high-level strategic intelligence.

Every intelligence assessment. Every threat surface. Every hardened deployment. The OPSEC discipline that keeps the architecture defensible from day one.

Sully Regmi

Sully Regmi

Research Scientist

Neuroscience and emotion-science researcher. Holds a degree in neuroscience. Studies how emotion and social judgment arise from neurocognitive processes at UC Santa Barbara's Yu Emotion Science Lab, with earlier work on visual attention and eye-tracking at UCSB's Vision and Image Understanding Lab. Brings that grounding in how the brain processes signals together with a machine-learning practice across natural language processing and computer vision.

On the OSCEN team, contributes on the speech and language side, connecting how biological auditory and speech-motor pathways learn to the platform's spiking implementation, and supporting the research program and experimental validation.

Research focus: natural language processing, computer vision, emotion and affective modeling, and reproducible evaluation of the learning system.

Investor inquiry

Start the conversation.

Tell us who you are. Accredited investors can book a 30-minute intro call right after sending this, and receive our deck under NDA. Everyone else hears back within 48 hours.

Are you an accredited investor in the US? (what's this?)

Nothing on this page constitutes an offer to sell or a solicitation to buy any security. All conversations are subject to applicable securities law.

← Back to overview
Chief Digital and Artificial Intelligence Office, Awardable, Tradewinds Solutions Marketplace

Privacy Terms