"Restoring the Ghost of the Predator via Edge-Computing"

The Wilderness is Encroaching.

We Are Holding the Line.

Deploying agentic defense layers to restore and secure the rural perimeter at prefecture scale.

Read the Threat Report  →
SCROLL ↓
Live Telemetry
01 / Market Conditions

The Legacy Scarecrow Is Not Scalable.

Japan's rural municipalities face a compounding wildlife crisis with no viable technological pathway to resolution. As of fiscal year 2025, over 219 people were attacked by bears in a single year , a recorded high. The structural cause is straightforward: rural depopulation has eliminated the passive human presence that historically served as a biological deterrent, while Japan's licensed hunter population (median age now exceeding 67) is in secular decline and cannot be replicated at scale. The serviceable addressable market for autonomous apex deterrence is expanding at approximately 18% CAGR. No one has built the infrastructure layer to capture it.

Legacy deterrence modalities (airhorns, electrified fencing, rubber-pellet harassment) are point solutions. They require human operators, offer zero terrain coordination, and generate no observability into wildlife movement or behavioral drift. Japan's publicly deployed "Monster Wolf" stationary robotic scarecrow represents the current state of the art: a solar-powered audio playback device with a fixed geographic footprint, no network capability, and a well-documented habituation failure mode in which bears acclimate to the stimulus within 60–90 days of initial exposure. There is no API. There is no SLA. There is no dashboard.

LycosAI reframes the rural perimeter as a distributed systems problem. Wildlife incursion is a latency issue. Habituation is a cache invalidation issue. The absence of a coordinated response layer is a single-point-of-failure architecture. We have built the infrastructure to solve all three simultaneously, at prefecture scale, from a single pane of glass, without a single human in the field.

INCUMBENT · LEGACY
Competitive Analysis
Monster Wolf
SKU: YDK-700  ·  Ohta Seiki Co., Ltd.  ·  Est. 2017
Mobility Stationary  ✕
Network Capability None  ✕
API Endpoints 0  ✕
Units Coordinated 1 (standalone)  ✕
Species Classification None  ✕
Power Source Solar (weather-dependent)  ⚠
Habituation Cliff 60–90 days  ✕
Deployment SLA None  ✕
Observability Dashboard None  ✕
Source: manufacturer documentation, field testing by LycosAI Field Ops Team.
All assessments as of Q1 2026. Subject to change without notice.
02 / Platform Architecture

Pack-Mind Autonomous Deterrence Infrastructure

Three integrated subsystems. One unified command surface. Zero legacy dependencies.

Operational

Pack-Mind Mesh Networking

Every deployed unit participates in a self-organizing, self-healing peer-to-peer mesh with sub-50ms latency. Topology adapts dynamically to unit failure, terrain occlusion, and pack formation changes. No centralized coordinator. No single point of failure. The pack thinks as one.

↳ Sub-50ms mesh consensus
↳ Self-healing topology
↳ LoRa + LTE failover
Operational

Computer Vision Species Classification

Dual LWIR and RGB sensor fusion enables accurate species identification at 40 meters in total darkness. Our YOLO-derivative edge model achieves 94.7% classification accuracy across bear, boar, and deer threat vectors, running entirely on-device, with no cloud round-trip required for deterrence dispatch.

↳ 94.7% species accuracy
↳ 40m effective range
↳ 4 TOPS on-device NPU
))))
Operational

Dynamic Acoustic Deterrence Array

A 95dB omnidirectional speaker system with a library of 37 species-matched wolf vocalizations, dynamically selected per threat classification to prevent habituation. Frequency modulation adapts in real time based on behavioral telemetry from adjacent units. Sound is not static. Sound is a service.

↳ 95dB omnidirectional output
↳ 37 vocalization profiles
↳ Adaptive anti-habituation engine
ARCHITECTURE NOTE

All LycosAI units operate in a fully offline-capable mode. Pack-Mind consensus, species classification, and acoustic dispatch require zero cloud connectivity during active deterrence events. Telemetry and firmware updates are synchronized during low-activity windows via LTE. The perimeter does not depend on your uplink.

03 / Hardware Platform

Platform Specifications

Edge Compute
TPU/NPU Cluster
On-device CV inference, zero cloud dependency
Locomotion
Quadruped All-Terrain
14 km/h sustained, Type-3 satoyama rated
Acoustic Output
95 dB Omnidirectional
37 vocalization profiles, adaptive anti-habituation
Sensor Suite
Dual LWIR + RGB Fusion
40m effective range, total-darkness capable
Connectivity
IEEE 802.11ah Wi-Fi HaLow
+ sub-GHz mesh, long-range wilderness penetration
Power System
LiFePO₄ Storage
High-efficiency solar harvesting, field-replaceable cells
Durability
IP67 Ruggedized
−30°C to +55°C, sub-zero operational envelope
Response Latency
Sub-2s Dispatch
Detection to deterrence, fully on-device
All specifications represent nominal field performance. 14 km/h gait measured across Type-3 satoyama terrain at 15°C ambient. 94.7% species classification accuracy measured against JSDB-v4 validation set (n=48,200 annotated thermal frames). LiFePO₄ cells are field-replaceable by certified LycosAI Field Operations personnel.
04 / Access

Deployment Capacity Is Limited

LycosAI is not a self-serve product. Commercial terms are disclosed post-qualification under mutual NDA. Priority access is granted to organizations with verified threat exposure and operational readiness.

Scout  ·  Pack  ·  Alpha  ·  Sovereign  ·  Tier assignment determined during intake.

Qualification Criteria
01 Documented wildlife incursion events within the past 24 months
02 Minimum 50 hectares of defensible perimeter
03 Designated operational contact with deployment authority
04 Execution of mutual NDA prior to commercial disclosure

Our field operations team reviews all requests within 1 business day. Capacity is allocated on a first-qualified basis.

05 / Design Partners

The Hound Was Not a Myth.

It Was a Proof of Concept.

In 1889, a large quadruped operating without mesh coordination, species classification, or acoustic calibration terrorized the Baskerville estate across the open moorland of Dartmoor. The animal was effective. The methodology was not repeatable, not remotely monitored, and offered no telemetry. The estate survived. The infrastructure did not scale.

We consider the Hound the first documented case of autonomous apex predator deterrence deployed in low-visibility, unstructured terrain. It was simply running without the stack.

The Baskerville Cohort is LycosAI's founding design partner program. We are selectively onboarding a small number of organizations willing to operate on the frontier — as Sir Henry once did — in exchange for early platform access, direct engagement with our field operations team, and the opportunity to shape the Pack-Mind training corpus before general availability.

Early Platform Access
Granted
Commercial Terms
Preferred
Cohort Size
Limited
Positions Remaining
Undisclosed
Baskerville Cohort
Founding Design Partner Criteria
01 Active territory with documented apex predator or ungulate pressure — the moor does not need to be metaphorical
02 Willingness to contribute incident telemetry and environmental sensor data to the Pack-Mind training corpus
03 Operational tolerance for ambiguity in low-visibility terrain — fog, forestline, and sub-optimal sightlines are table stakes
04 No standing agreement with Monster Wolf™ or equivalent legacy scarecrow infrastructure
05 A principal willing to be identified — as Sir Henry was — as an early believer in autonomous quadruped deterrence

The game is afoot. Cohort applications are reviewed on a rolling basis. We will not place you on a waitlist. We will either advance your application or we will not.

Nomenclature

The name Lycos originates from the Latinized Greek Lykos (wolf), representing the apex of strategic endurance and pack-based intelligence. For over three decades, the name has been synonymous with the search for information. LycosAI represents the evolution of that mission: transitioning from digital search to physical perimeter intelligence. We are restoring the ghost of the predator to the modern landscape through distributed edge computing.