From Fugaku to SLIM: An Exhaustive Analysis of Japan’s Integrated Cyber-Physical Strategy

Abstract

As the world navigates the mid-2020s, Japan has aggressively reasserted its position as a global leader in high-technology research and development. Driven by the "Society 5.0" initiative—a national strategy to integrate cyberspace and physical space to solve social problems—Japanese research institutions and private enterprises have achieved significant milestones between 2024 and early 2026. This report provides an exhaustive analysis of breakthroughs in four convergent domains: aerospace engineering, high-performance computing, artificial intelligence, and advanced robotics.

The analysis reveals a distinct "Japanese Model" of innovation characterized by the hybridization of classical and novel technologies (such as bio-hybrid robotics and CPU-GPU supercomputing), a strategic emphasis on sovereign capabilities (domestic LLMs and launch vehicles), and a relentless focus on precision engineering (pinpoint lunar landings and quantum reliability). While the period has been marked by high-profile successes, including the operationalization of a 256-qubit quantum computer and the historic SLIM lunar landing, it has also faced setbacks in commercial spaceflight and launch infrastructure. This report synthesizes these events into a cohesive narrative, offering technical depth on the mechanisms driving these innovations and their broader implications for the future of science.

1. Aerospace Engineering: Precision, Autonomy, and Sustainability in Japan

The Japanese aerospace sector has undergone a fundamental transformation, shifting from a focus on heavy-lift reliability to precision exploration and orbital sustainability. This era is defined by the mastery of autonomous navigation systems capable of operating in gravity wells and the commercialization of orbital maintenance.

1.1 The SLIM Paradigm: Mastery of Pinpoint Lunar Landing

In January 2024, the Japan Aerospace Exploration Agency (JAXA) achieved a historic milestone with the Smart Lander for Investigating Moon (SLIM), making Japan the fifth nation to soft-land on the lunar surface. However, the significance of SLIM lies not in the landing itself, but in the precision with which it was executed, transitioning the paradigm of planetary exploration from "landing where it is easy" to "landing where we want".¹

1.1.1 Vision-Based Navigation and Autonomous Guidance

The core technological breakthrough of SLIM was its "Smart Eyes" vision-based navigation system. Traditional lunar landers have typically relied on radio guidance and rough altimetry, resulting in landing ellipses often spanning several kilometers. In contrast, SLIM utilized onboard image processing algorithms to estimate its position with unprecedented accuracy in real-time.¹

The guidance and navigation control (GNC) system operated in two critical phases:

1. Orbital Descent Initiation: Before beginning its powered descent, the spacecraft matched features observed by its navigation cameras against a pre-loaded, high-resolution map of the lunar surface derived from previous missions like Kaguya (SELENE) and the Lunar Reconnaissance Orbiter (LRO). This allowed for precise estimation of the spacecraft's position relative to the target site.¹

2. Vertical Descent and Obstacle Avoidance: During the vertical descent phase, at an altitude of approximately 50 meters, the lander shifted priority to obstacle detection. The onboard system scanned the surface for hazardous boulders and slopes, autonomously adjusting its lateral position to identifying the safest landing spot within the target zone. This capability is critical for future missions to the lunar poles, where safe landing zones near water ice deposits may be only a few hundred meters wide.¹

1.1.2 The "Two-Step" Landing Dynamics

The target landing site near the Shioli crater presented a unique engineering challenge: a significant slope of approximately 15 degrees. A conventional four-legged landing gear, designed for flat terrain, would risk toppling over upon impact. To mitigate this, JAXA engineers developed a novel "two-step" landing method.³

The mechanism was designed to dissipate kinetic energy through a controlled tumbling motion:

1. Main Gear Contact: The lander was configured to touch down first with its main landing gear, located on the rear of the spacecraft relative to its upright orientation.

2. Forward Rotation: Upon impact, the momentum would cause the spacecraft to tip forward, allowing the front auxiliary legs to settle onto the surface, stabilizing the lander against the slope.

While the actual landing resulted in the spacecraft resting in an unintended nose-down orientation due to a nozzle failure in one of the two main engines just prior to touchdown, the primary objective of "pinpoint" accuracy was achieved. The landing occurred approximately 55 meters east of the target point, a deviation orders of magnitude smaller than the kilometers-wide margins of previous international missions.¹ This validated the vision-based navigation algorithms that will be inherited by the upcoming Martian Moon eXploration (MMX) mission.⁴

1.2 Commercial Lunar Exploration: The Evolution of HAKUTO-R

Parallel to JAXA’s scientific endeavors, the Japanese private sector, led by ispace, has aggressively pursued commercial lunar access through the HAKUTO-R program. The trajectory of ispace demonstrates the iterative and high-risk nature of commercial spaceflight.

1.2.1 Mission 2 "RESILIENCE": Technical Analysis of the Landing Anomaly

Following the failure of Mission 1 in 2023, which was caused by a software logic error regarding altitude estimation near a crater rim, ispace launched Mission 2 in January 2025. The mission featured the "RESILIENCE" lander and aimed to deploy the "TENACIOUS" micro-rover.⁵

On June 6, 2025, the lander attempted its descent. Unlike Mission 1, where the propellant was exhausted due to a hovering holding pattern caused by conflicting sensor data, Mission 2 failed due to a specific hardware anomaly in the final moments. Post-flight analysis revealed that while the lander's attitude was nominal (vertical) during the final descent, the Laser Range Finder (LRF) failed to provide valid distance measurements to the onboard computer.⁷

Table 1: Comparative Analysis of HAKUTO-R Mission Anomalies

Feature	Mission 1 (2023)	Mission 2 (2025)
Lander Identification	Series 1 (Unnamed)	Series 1 (RESILIENCE)
Failure Phase	Terminal Descent (Hover)	Terminal Descent (Braking)
Primary Anomaly	Software Logic: The terrain relative navigation filter rejected valid laser altimeter data as "noise" because of a sudden altitude change when passing over a crater rim.	Hardware Sensor: The Laser Range Finder (LRF) experienced internal delays or failure, preventing the computer from calculating the correct time to fire braking thrusters.
Impact Velocity	Freefall from ~5km altitude after fuel exhaustion.	High-velocity impact due to insufficient deceleration.
Key Lesson	Necessity of robust terrain filtering logic for complex topography.	Criticality of sensor hardware resilience in deep space environments.

Despite the hard landing, Mission 2 successfully validated the deep-space orbital injection and navigation maneuvers, confirming the company's ability to deliver payloads to lunar orbit.⁹ ispace has since constituted an external review task force and is proceeding with the development of Mission 3 (APEX 1.0 lander) scheduled for 2026, which will be led by its U.S. subsidiary.⁷

1.3 Launch Vehicle Sovereignty: The H3 Rocket Program

The H3 launch vehicle, developed jointly by JAXA and Mitsubishi Heavy Industries, represents Japan's bid to maintain autonomous access to space with a cost-effective heavy-lift rocket. The program has faced a volatile development history, oscillating between success and failure in the 2024–2026 window.

1.3.1 Operational Volatility (2024–2025)

The H3 program stabilized temporarily in 2024 with the success of Flight 2 (February) and Flight 3 (July), validating the performance of the LE-9 first-stage engine. The LE-9 utilizes an "expander bleed cycle," a sophisticated thermodynamic cycle that drives the turbopumps using heat from the combustion chamber cooling channels. This design is theoretically safer and more efficient than gas-generator cycles but historically difficult to scale to high thrust.¹¹

However, the program faced a critical setback with Flight 8 on December 22, 2025.

1.3.2 Flight 8 Investigation: The Second Stage Vulnerability

Flight 8 was tasked with deploying QZS-5 (Michibiki-5), a satellite vital for the Quasi-Zenith Satellite System (QZSS)—Japan's regional satellite positioning system that augments GPS signals for centimeter-level accuracy.¹²

Approximately 27 minutes after liftoff, the second-stage engine (LE-5B-3) failed to execute its planned second ignition. The engine shut down prematurely, stranding the payload in a non-viable orbit. This marked the second time in the H3's short history that the second stage was the point of failure (the first being the inaugural flight in 2023).¹²

The failure has significant strategic implications. The QZSS constellation is the backbone of Japan's autonomous logistics and drone infrastructure (discussed in Section 4). The loss of QZS-5 delays the constellation's completion, forcing JAXA and the government to establish a Special Task Force to re-evaluate the reliability of the upper stage propulsion electrical systems before the next launch.¹³

1.4 Orbital Robotics and Debris Remediation

Japan has emerged as a global leader in the niche but critical field of orbital sustainability, driven by the commercial innovations of Astroscale and Gitai.

1.4.1 Astroscale: From Inspection to Removal

Astroscale’s ADRAS-J (Active Debris Removal by Astroscale-Japan) mission achieved a major technical breakthrough in 2024. The spacecraft successfully performed Rendezvous and Proximity Operations (RPO) around a non-cooperative target—a spent H-IIA upper stage rocket body that had been drifting in orbit since 2009.¹⁵

Navigating around a "non-cooperative" object (one that does not transmit GPS data or communicate) requires advanced autonomy. ADRAS-J utilized a suite of sensors to perform "Angles Only Navigation," eventually transitioning to "Model Matching Navigation," where the onboard computer matched the visual data of the rocket body with a 3D model to determine its tumble rate and spin axis. The spacecraft successfully maintained a fixed distance of approximately 50 meters, capturing high-resolution diagnostic images.¹⁵

Moving into 2025 and 2026, Astroscale has begun development of ADRAS-J2 and ISSA-J1. Unlike the first mission, which was purely observational, ADRAS-J2 is designed to capture the debris using a robotic arm and actively de-orbit it, marking the transition from debris monitoring to active environmental remediation.¹⁷

1.4.2 Gitai: The Autonomous Space Workforce

Gitai has expanded its robotic capabilities from the International Space Station (ISS) to independent satellite platforms. In 2024, the company’s S2 dual-robotic arm system successfully executed assembly and maintenance tasks outside the ISS Bishop Airlock, achieving NASA Technology Readiness Level (TRL) 7.¹⁹

Following this, Gitai launched a 16U-sized satellite in late 2024 (operational in early 2025) equipped with robotic arms. This mission demonstrated the capability to perform end-to-end satellite servicing tasks autonomously, positioning the company to begin commercial on-orbit servicing by 2026.²⁰

2. Next-Generation Computing: The Hybridization of Silicon and Qubits

Japan’s computing strategy has shifted from a singular focus on raw FLOPs (floating-point operations per second) to a hybridized architecture that integrates High-Performance Computing (HPC) with AI acceleration and quantum processing. This shift acknowledges the "AI for Science" paradigm, where simulation and machine learning are inextricably linked.

2.1 The Post-Fugaku Era: "FugakuNEXT"

The supercomputer Fugaku, which held the title of the world's fastest supercomputer from 2020 to 2022, is being succeeded by a new system provisionally named FugakuNEXT. Initiated in 2025 by RIKEN and Fujitsu, this system targets deployment around 2030 and represents a significant architectural departure from its predecessor.²¹

2.1.1 Zetta-Scale AI Performance

While Fugaku was an Exascale machine, FugakuNEXT is designed to achieve Zetta-scale performance specifically for AI workloads. The target is to exceed 600 ExaFLOPS (0.6 ZettaFLOPS) in FP8 precision, a low-precision format optimized for training massive neural networks. This capability is intended to support the training of trillion-parameter AI models that are currently computationally prohibitive for scientific discovery.²²

2.1.2 Heterogeneous Architecture: Monaka Meets NVIDIA

Unlike Fugaku, which utilized a homogeneous architecture based solely on the Fujitsu A64FX CPU, FugakuNEXT will employ a heterogeneous design.

• FUJITSU-MONAKA (Monaka-X): The central processor is a next-generation Arm-based CPU built on a 2nm process node. It features a 3D many-core architecture with 144 cores per socket (288 per node) and utilizes ultra-low voltage technology to maximize energy efficiency.²⁴

• NVLink Fusion: In a strategic collaboration, the Monaka CPUs will be paired with NVIDIA GPUs. The system will utilize NVLink Fusion to create a high-bandwidth, coherent interconnect between the CPU and GPU. This allows for unified memory addressing, enabling the CPU to handle complex physics simulations while offloading matrix-heavy AI tasks to the GPU without the latency bottlenecks typical of PCIe connections.²⁶

2.2 Quantum Computing: Scaling the Superconducting Qubit

Japan has accelerated its domestic quantum roadmap, aiming to bridge the gap between Noisy Intermediate-Scale Quantum (NISQ) devices and Fault-Tolerant Quantum Computing (FTQC).

2.2.1 From 64 to 256 Qubits (2025)

In April 2025, the RIKEN-Fujitsu collaboration unveiled a 256-qubit superconducting quantum computer, a quadrupling of capacity from the 64-qubit system launched in 2023.²⁸

The primary engineering challenge in scaling superconducting qubits is the "wiring bottleneck." Qubits must be controlled by microwave signals sent from room temperature electronics down to the millikelvin environment of the dilution refrigerator. As qubit counts increase, the number of cables becomes unmanageable, introducing heat and occupying limited volume.

• Solution: The 256-qubit system utilizes a novel 3D packaging and connection structure. This architecture allows signal wires to be routed vertically through the chip stack rather than laterally, enabling high-density integration without compromising the thermal isolation required to maintain quantum coherence.²⁸

2.2.2 The Hybrid Platform Strategy

Japan’s strategy involves the immediate integration of these quantum resources into a hybrid platform. By connecting the quantum computer to HPC resources (like Fugaku), researchers can execute hybrid algorithms such as the Variational Quantum Eigensolver (VQE). In this setup, the quantum computer calculates the energy states of a molecule (a task difficult for classical computers), while the supercomputer handles the parameter optimization loop. This "ABC" strategy (AI, Big Data, Computing) aims to make quantum computing practically useful for material science before full fault tolerance is achieved.²⁹

2.3 Quantum-Inspired Optimization: The Simulated Bifurcation Machine

While true quantum computers mature, Toshiba has commercialized the Simulated Bifurcation Machine (SBM), a "quantum-inspired" technology that solves combinatorial optimization problems using classical hardware.

The SBM operates by simulating a physical phenomenon known as bifurcation. It maps a discrete optimization problem (typically an Ising model with binary variables) onto a system of continuous non-linear equations. As a control parameter is evolved, the system "bifurcates" into stable oscillatory states that correspond to the optimal solution of the problem.³¹

In 2025, this technology was successfully applied to real-time dynamic environments, such as "Collaborative Filtering" for recommendation engines and "Black-Box Optimization" for industrial control systems. The SBM has demonstrated speedups of up to 100 times compared to traditional simulated annealing methods, proving that quantum physics principles can deliver value on classical silicon today.³²

3. Artificial Intelligence: Sovereignty and Evolution

Japan's approach to Artificial Intelligence is distinct from the US and Chinese models. It emphasizes "sovereignty" (domestic control over models and data), "cultural nuance" (mastery of the Japanese language and high-context communication), and "efficiency" (achieving results with lower computational and energy costs).

3.1 The Rise of Sovereign Large Language Models (LLMs)

Recognizing the risks of relying on foreign "black box" models, Japanese research institutes and corporations have prioritized the development of domestic LLMs.

3.1.1 Fugaku-LLM: The Transparent Standard

Released in May 2024, Fugaku-LLM is a 13-billion parameter model trained entirely on the Fugaku supercomputer. Its development was a proof-of-concept for training Transformers on CPU-based architectures. Researchers ported the Megatron-DeepSpeed deep learning framework to Fugaku’s A64FX architecture, optimizing the dense matrix multiplication libraries to run efficiently on CPUs connected via the Tofu Interconnect D network.³⁴

The model achieved top-tier performance on the Japanese MT-Bench, particularly in humanities and social sciences tasks, outperforming open-source models of similar size. More importantly, it is a "transparent" model—trained on proprietary Japanese data with fully disclosed architecture—allowing for safe commercial and academic use where data provenance is critical.³⁵

3.1.2 Corporate Sovereignty: Sarashina and tsuzumi

In the private sector, SoftBank and NTT have launched models tailored to industrial security.

• SoftBank (Sarashina): In October 2025, SoftBank evolved its "Large Telecom Model" (LTM) by integrating Sarashina, a homegrown LLM specialized in Japanese business and legal contexts. This integration ensures that sensitive telecommunications data is processed entirely within Japan, adhering to strict data governance laws.³⁶

• NTT (tsuzumi): Launched in late 2025, tsuzumi is a lightweight LLM designed for energy efficiency. Unlike massive models that require hyperscale data centers, tsuzumi can run in local environments (on-premise), making it ideal for the medical and financial sectors where data cannot leave the facility.³⁸

3.2 Nature-Inspired AI: Sakana AI

A standout in the Japanese startup ecosystem is Sakana AI, a Tokyo-based company founded by former Google researchers. Sakana AI challenges the prevailing "scaling laws" (which advocate for simply adding more compute and data) by employing nature-inspired evolutionary algorithms.

3.2.1 Evolutionary Model Merge (EMM)

Sakana AI developed a technique to create new foundation models by "breeding" existing ones, rather than training them from scratch. This process, known as Evolutionary Model Merge, operates in both parameter space (model weights) and data-flow space (layer architecture).

Using evolutionary algorithms, the system automatically identifies the optimal combination of layers from different open-source models—for instance, merging a model with strong mathematical reasoning with one possessing high Japanese linguistic fluency. The resulting models, such as EvoLLM-JP and EvoVLM-JP (Vision-Language), achieved state-of-the-art performance on Japanese benchmarks with a fraction of the compute required for traditional training.³⁹

3.2.2 The AI Scientist: Promise and Pitfalls

In August 2024, Sakana AI introduced "The AI Scientist," an ambitious system designed to automate the entire scientific research lifecycle. The system autonomously generates research ideas, writes code to test them, executes experiments, drafts a paper in LaTeX, and even performs a simulated peer review.⁴¹

While the system successfully produced papers accepted at simulated conferences, independent evaluations in 2025 revealed significant limitations. The system exhibited a code execution failure rate of approximately 42%, often generated hallucinated citations, and struggled with true novelty.⁴² However, the project remains a pioneering attempt at "Artificial Research Intelligence," suggesting a future where AI acts not just as a tool, but as a collaborator in the scientific method.

3.3 AI for Science: The Closed-Loop Laboratory

The National Institute for Materials Science (NIMS) has pioneered the integration of AI into the physical workflow of material discovery, creating "Self-Driving Laboratories."

3.3.1 NIMS-OS (Orchestration System)

NIMS-OS is a middleware platform that connects AI algorithms (the "brain") with robotic synthesis equipment (the "hands"). The system creates a closed loop: the AI (typically using Bayesian optimization) suggests a material composition; the robot synthesizes the material and measures its properties; and the data is fed back to the AI to refine the next suggestion. This cycle runs continuously without human intervention.⁴³

In 2025, researchers utilized NIMS-OS to achieve a breakthrough in battery electrolytes. The system autonomously navigated the complex chemical space of polymers and salts to discover a new solid polymer electrolyte for Lithium-metal batteries. The AI identified compositions with high ionic conductivity that human researchers had overlooked, reducing the discovery timeline from an estimated six years to just one month.⁴⁴

4. Advanced Robotics: From Biological Fusion to Avatar Symbiosis

Japanese robotics is evolving beyond the factory floor. The current wave of innovation focuses on bio-hybrid systems that integrate living tissue, and cybernetic avatars that extend human presence, driven by the pressing needs of an aging society.

4.1 Bio-Hybrid Robotics: The Fusion of Flesh and Machine

At the University of Tokyo, Professor Shoji Takeuchi’s Biohybrid Systems Laboratory is blurring the line between biology and engineering by incorporating living tissue into robotic structures.

4.1.1 The "Sushi-Roll" Muscle Actuator (MuMuTA)

One of the primary challenges in bio-hybrid robotics is keeping muscle tissue alive and functional; thick muscle bundles typically undergo necrosis (cell death) due to a lack of nutrient perfusion. In February 2025, Takeuchi’s team published details of a breakthrough solution: the Multiple Muscle Tissue Actuator (MuMuTA).

The team aligned myoblasts (muscle precursor cells) within a collagen gel and rolled them into thin, thread-like bundles, resembling sushi rolls. This geometry maximized the surface area exposed to the nutrient culture medium, allowing the tissue to survive and mature. By bundling these threads together, they created a powerful linear actuator capable of exerting significant force.⁴⁶

To demonstrate this, the team built a bio-hybrid robot hand. The MuMuTAs acted as antagonists (like biceps and triceps), allowing the robot fingers to perform complex motions, such as the "scissor" gesture in rock-paper-scissors, upon electrical stimulation. This research paves the way for prosthetics that move with the natural fluidity of biological organisms.⁴⁷

4.1.2 Engineered Living Skin

In a related development, the same laboratory succeeded in attaching engineered living skin to a robotic face. A key innovation was the use of V-shaped perforations in the robot's underlying structure, designed to mimic the function of human ligaments. These anchors allowed the living skin to attach securely to the mechanical frame, enabling it to stretch and deform with the robot's facial expressions without tearing or peeling. This advancement brings the field closer to creating robots with the self-healing and tactile capabilities of human skin.⁴⁹

4.2 Moonshot Goal 1: The Avatar Symbiotic Society

The Japanese government’s "Moonshot R&D Program" has set an ambitious goal (Goal 1): to realize a society in which human beings can be free from limitations of body, brain, space, and time by 2050.

4.2.1 Cybernetic Avatars and the Internet of Brains

Led by renowned roboticist Hiroshi Ishiguro, this project focuses on the development of Cybernetic Avatars (CA)—robotic or digital bodies that humans can operate remotely. The vision is to allow anyone, regardless of physical ability or location, to participate in the workforce.

In November 2025, the project conducted a major demonstration in Dubai, showcasing the "Internet of Brains." This technology utilizes Brain-Machine Interfaces (BMI) to allow operators to control avatars using only brain signals. This capability is specifically targeted at empowering individuals with paralysis or severe physical disabilities to engage in social and economic activities.⁵⁰

4.2.2 Robotics for an Aging Society

With Japan’s population aging rapidly and a projected shortage of 570,000 caregivers by 2040, robotics is seen as an essential infrastructure.

• AIREC: Developed by Waseda University, AIREC is a humanoid robot designed to assist with physically demanding care tasks, such as transferring patients from beds to wheelchairs and assisting with continence care.⁵²

• Diagnostic Robotics: Fujitsu has deployed aiGait, a system that uses AI to analyze the walking patterns (gait) of elderly patients. By monitoring subtle changes in stride and posture, the system can detect early signs of cognitive decline or motor disorders, allowing for preventative care.⁵²

4.3 Disaster Response Robotics

The Noto Peninsula earthquake and other natural disasters have accelerated the deployment of field robotics in Japan.

• Q-UGVs: The Japan Ground Self-Defense Force has begun deploying quadrupedal Unmanned Ground Vehicles (Q-UGVs) for reconnaissance. These robots can navigate landslide zones and unstable rubble that are too dangerous for human responders, delivering supplies and gathering data.⁵³

• SMURF: In collaboration with EU researchers, Japanese teams developed the SMURF (Soft Miniaturized Underground Robotic Finder). This compact, two-wheeled robot is designed to scramble deep into debris piles to locate survivors using chemical sensors and cameras, communicating its findings to a mothership drone.⁵⁴

• Dragon Firefighter: Tohoku University developed a unique "flying hose" robot. Guided by water jets, the dragon-like robot can fly directly into high-rise windows to extinguish fires at the source, navigating obstacles that would block traditional aerial ladders.⁵⁵

5. Conclusion

The scientific and engineering developments of 2024–2026 illustrate a Japan that is actively re-engineering its technological foundations. The "Japanese Model" emerging from this period is defined by hybridization—the seamless integration of biological tissue with mechanics, of quantum qubits with supercomputing nodes, and of human operators with cybernetic avatars.

Furthermore, there is a clear strategic imperative for sovereignty. The development of the H3 rocket, domestic LLMs like Sarashina, and homegrown quantum computers reflects a determination to maintain national autonomy in critical infrastructure. Yet, this technology is deeply aligned with social utility. From the precise lunar landings of SLIM enabling future resource extraction to the avatar robots of Moonshot Goal 1 addressing the labor crisis, Japanese innovation remains rooted in the specific, tangible needs of its society. As these technologies mature, they offer a blueprint for how advanced nations can navigate the complexities of the 21st century through the fusion of human intent and machine capability.

Works cited

1. Smart Lander for Investigating Moon (SLIM) Results from the Moon Landing Institute of Space and Astronautical Science SLIM Proje - JAXA, accessed January 8, 2026, https://global.jaxa.jp/press/2024/01/file/jaxa_doc01_20240125_e.pdf

2. SLIM Project Press Kit - JAXA, accessed January 8, 2026, https://global.jaxa.jp/countdown/slim/SLIM-mediakit-EN_2308.pdf

3. 20 minutes of terror: SLIM will attempt a pinpoint accurate landing on the lunar surface, accessed January 8, 2026, https://cosmos.isas.jaxa.jp/20-minutes-of-terror-slim-will-attempt-a-pin-point-accurate-landing-on-the-lunar-surface/

4. SLIM, Japan's precision lunar lander | The Planetary Society, accessed January 8, 2026, https://www.planetary.org/space-missions/slim-japans-precision-lunar-lander

5. ispace Announces Mission 2 Launch Date & Time, accessed January 8, 2026, https://ispace-inc.com/news-en/?p=6678

6. Press Release January 9, 2025 ispace, inc. ispace Announces Mission 2 Launch Date & Time RESILIENCE Lunar Lander &a, accessed January 8, 2026, https://ispace-inc.com/wp-content/uploads/2025/01/20250109_M2-Launch-Date.pdf

7. ispace Releases Technical Cause Analysis for HAKUTO-R Mission 2, accessed January 8, 2026, https://ispace-inc.com/news-en/?p=7671

8. Status Update on ispace Mission 2 SMBC x HAKUTO-R Venture Moon, accessed January 8, 2026, https://ispace-inc.com/news-en/?p=7664

9. ispace Completes Success 7 of Mission 2 Milestones, accessed January 8, 2026, https://ispace-inc.com/news-en/?p=7327

10. ispace Announces Mission 2 Landing Date Set for June 6, 2025 (JST), accessed January 8, 2026, https://ispace-inc.com/news-en/?p=7097

11. H3 (rocket) - Wikipedia, accessed January 8, 2026, https://en.wikipedia.org/wiki/H3_(rocket)

12. H3 rocket suffers upper stage anomaly, fails to correctly deploy navigation satellite, accessed January 8, 2026, https://spaceflightnow.com/2025/12/22/h3-rocket-suffers-upper-stage-anomaly-fails-to-correctly-deploy-navigation-satellite/

13. Japan loses another H3 launcher, plus the satnav bird it carried, accessed January 8, 2026, https://www.theregister.com/2025/12/23/jaxa_h3_failure_inquiry/

14. Launch Failure of the 8th H3 Launch Vehicle and Setting up a Special Task Force - JAXA, accessed January 8, 2026, https://global.jaxa.jp/press/2025/12/20251222-2_e.html

15. CRD2 Phase I / ADRAS-J Update: Fly-Around Observation Images of Space Debris Released - Japan Aerospace Exploration Agency, accessed January 8, 2026, https://global.jaxa.jp/press/2024/07/20240730-1_e.html

16. Astroscale's ADRAS-J Achieves Historic 15-Meter Approach to Space Debris, accessed January 8, 2026, https://www.astroscale.com/news/astroscales-adras-j-achieves-historic-15-meter-approach-to-space-debris

17. Astroscale Japan Selected for Phase II of JAXA's Commercial Removal of Debris Demonstration Program, accessed January 8, 2026, https://www.astroscale.com/en/news/astroscale-japan-selected-for-phase-ii-of-jaxas-commercial-removal-of-debris-demonstration-program

18. Astroscale Japan Transitions to Phase II of Satellite Debris Inspection Mission Development, accessed January 8, 2026, https://www.astroscale.com/en/news/astroscale-japan-transitions-to-phase-ii-of-satellite-debris-inspection-mission-development

19. GITAI Completes Fully Successful Technology Demonstration Outside the ISS, accessed January 8, 2026, https://gitai.tech/2024/03/19/gitai-completes-fully-successful-technology-demonstration-outside-the-iss/

20. GITAI Successfully Completes Space Demonstration of In-House Developed 16U Satellite, accessed January 8, 2026, https://gitai.tech/2025/01/27/gitai-successfully-completes-space-demonstration-of-in-house-developed-16u-satellite/

21. Fujitsu awarded contract to design next-generation flagship supercomputer FugakuNEXT, accessed January 8, 2026, https://www.fujitsu.com/global/about/resources/news/press-releases/2025/0618-01.html

22. The Convergence of HPC, AI, and Quantum: Towards FugakuNEXT “made with Japan”, accessed January 8, 2026, https://www.hpcuserforum.com/wp-content/uploads/2025/10/Satoshi-Matsuoka_The-Convergence-of-HPC-AI-and-Quantum-FugakuNEXT_Oct-2025-HPC-UF.pdf

23. RIKEN launches international initiative with Fujitsu and NVIDIA for "FugakuNEXT" development - 理化学研究所, accessed January 8, 2026, https://www.riken.jp/en/news_pubs/news/2025/20250822_1/index.html

24. Next Arm Processor FUJITSU-MONAKA and Its Technologies, accessed January 8, 2026, https://www.fujitsu.com/global/imagesgig5/FUJITSU-MONAKA.pdf

25. FUJITSU-MONAKA series: Arm-based processor, accessed January 8, 2026, https://global.fujitsu/-/media/Project/Fujitsu/Fujitsu-HQ/technology/research/events/sc25/SC25-FUJITSU-MONAKA.pdf?rev=138be9d7353849f3982e64617b347e20&hash=10F0FFEC2C6D5B703416334F3889D409

26. RIKEN, Japan's Leading Science Institute, Taps Fujitsu and NVIDIA for Next Flagship Supercomputer, accessed January 8, 2026, https://blogs.nvidia.com/blog/fugakunext/

27. FugakuNEXT Project: next Japanese flagship supercomputer - Indico Global, accessed January 8, 2026, https://indico.global/event/14504/contributions/137187/attachments/64281/124203/20251106_Lattice2025_yaoki_v1_pdf.pdf

28. Fujitsu and RIKEN develop world-leading 256-qubit superconducting quantum computer, accessed January 8, 2026, https://www.fujitsu.com/global/about/resources/news/press-releases/2025/0422-01.html

29. Fujitsu Quantum Computing Strategy, accessed January 8, 2026, https://www.fujitsu.com/tw/imagesgig5/20250422-01f.pdf

30. RIKEN and Fujitsu develop 64-qubit quantum computer, accessed January 8, 2026, https://www.riken.jp/en/news_pubs/news/2023/20231005_2/index.html

31. Technologies | Quantum-Inspired Optimization Solutions SQBM+ | TOSHIBA DIGITAL SOLUTIONS CORPORATION, accessed January 8, 2026, https://www.global.toshiba/ww/products-solutions/ai-iot/sbm/sbm-technologies.html

32. News and Topics | Quantum-Inspired Optimization Solutions SQBM+ | TOSHIBA DIGITAL SOLUTIONS CORPORATION, accessed January 8, 2026, https://www.global.toshiba/ww/products-solutions/ai-iot/sbm/news.html

33. Toshiba joins Azure Quantum, providing Ising machine to solve large combinatorial optimization problems, accessed January 8, 2026, https://azure.microsoft.com/en-us/blog/quantum/2020/09/22/toshiba-joins-azure-quantum-network-machine-solving-large-combinatorial-optimization-problems/

34. Release of “Fugaku-LLM” – a large language model trained on the supercomputer “Fugaku” - Fujitsu Uvance, accessed January 8, 2026, https://www.fujitsu.com/global/about/resources/news/press-releases/2024/0510-01.html

35. Release of "Fugaku-LLM" — a large language model trained on the supercomputer "Fugaku" - 東京工業大学, accessed January 8, 2026, https://www.titech.ac.jp/english/news/2024/069223

36. SoftBank Corp.'s Large Telecom Model—a Generative AI Foundation for the Telecom Industry—Evolved into a Domestic AI Model and Launched for Internal Use | About Us, accessed January 8, 2026, https://www.softbank.jp/en/corp/news/press/sbkk/2025/20251029_01/

37. SoftBank Corp.'s Large Telecom Model—a Generative AI Foundation for the Telecom Industry— Evolved into a Domestic AI Model and Launched for Internal Use | About Us, accessed January 8, 2026, https://www.softbank.jp/en/corp/technology/research/topics/173/

38. NTT's Next-Generation LLM "tsuzumi 2" Now Available ~Supporting Japan's corporate digital transformation with a high-performance, high-security, low-cost purely domestic LLM~ | Press Release - NTT Group, accessed January 8, 2026, https://group.ntt/en/newsrelease/2025/10/20/251020a.html

39. Letting nature lead: How Sakana AI is transforming model building | AWS Startups, accessed January 8, 2026, https://aws.amazon.com/startups/learn/letting-nature-lead-how-sakana-ai-is-transforming-model-building

40. Evolutionary Optimization of Model Merging Recipes - arXiv, accessed January 8, 2026, https://arxiv.org/html/2403.13187v1

41. The AI Scientist Due Dillidence Report - Intor with AI, accessed January 8, 2026, https://www.intor.ai/ai-analysis/the-ai-scientist

42. Evaluating Sakana's AI Scientist for Autonomous Research: Wishful Thinking or an Emerging Reality Towards 'Artificial Research Intelligence' (ARI)? - arXiv, accessed January 8, 2026, https://arxiv.org/html/2502.14297v2

43. Full article: NIMS-OS: an automation software to implement a closed loop between artificial intelligence and robotic experiments in materials science - Taylor & Francis Online, accessed January 8, 2026, https://www.tandfonline.com/doi/full/10.1080/27660400.2023.2232297

44. Aligning AI and Robotic System for Autonomous Automated Materials Explorations, accessed January 8, 2026, https://nimsnow.nims.go.jp/en/research/2024no3_r03/

45. Full article: Seamless integration of legacy robotic systems into a self-driving laboratory via NIMO: a case study on liquid handler automation - Taylor & Francis Online, accessed January 8, 2026, https://www.tandfonline.com/doi/full/10.1080/27660400.2025.2565144

46. Japanese research team develops world's largest 'biohybrid' robot hand - The Japan Times, accessed January 8, 2026, https://www.japantimes.co.jp/news/2025/02/13/japan/science-health/largest-biohybrid-robot-hand/

47. Japan's Shocking Breakthrough: Robot Hand with Real Human Muscles + UBTECH's Hyperrealistic AI Robot - YouTube, accessed January 8, 2026, https://www.youtube.com/watch?v=d77r_6tQvqo

48. Biohybrid hand gestures with human muscles - EurekAlert!, accessed January 8, 2026, https://www.eurekalert.org/news-releases/1072918

49. Robots face the future | The University of Tokyo, accessed January 8, 2026, https://www.u-tokyo.ac.jp/focus/en/press/z0508_00360.html

50. Moonshot Goal 1 Kanai Project “Internet of Brains” Commences Joint Research with Dubai Future Labs, Aiming for Breakthroughs in Multi-platform Robot/Avatar Control System through BMI Technology Application, accessed January 8, 2026, https://brains.link/en/news/4128

51. Kan Akutsu, R&D Department Leader at Araya, and colleagues will host a public event in Dubai: “From Mind to Motion: Exploring the Future of Bio-Signal Interfaces for Robots and Avatars” (Dubai, November 26–27, 2025) | 株式会社アラヤ, accessed January 8, 2026, https://www.araya.org/en/publications/event20251113/

52. Japan turns to AI and robotics to tackle dementia crisis | Digital Watch Observatory, accessed January 8, 2026, https://dig.watch/updates/japan-turns-to-ai-and-robotics-to-tackle-dementia-crisis

53. Autonomous Robots for Earthquake Disaster Relief | Case Studies, accessed January 8, 2026, https://www.ghostrobotics.io/case-studies/autonomous-robots-for-earthquake-disaster-relief

54. Robots to the rescue: miniature robots offer new hope for search and rescue operations | Horizon Magazine - Research and innovation, accessed January 8, 2026, https://projects.research-and-innovation.ec.europa.eu/en/horizon-magazine/robots-rescue-miniature-robots-offer-new-hope-search-and-rescue-operations

55. Bringing life-saving robotics from the disaster areas of Miyagi and Fukushima to the world. - 復興庁, accessed January 8, 2026, https://expo2025-portal.reconstruction.go.jp/english/bbb/tech/07/