#Jetson Nano Nvidia

[nullset2]

I'm a dumbass

I have a Turing Pi V2 with a Jetson Nano and it's the bees knees but I wanted to clear out some space, so I removed the graphical environment and other packages from Linux4Tegra (which is what NVIDIA offers, basically repackaged Ubuntu 18 LTS with their drivers and facilities, which have been historically NVIDIA's weak point in Linux land).

But the tutorial says "remove network-manager and then reinstall it", and I thought "well, I'll follow the tutorial", foolishly unaware that I had disabled dhcpd earlier in the week due to another issue, so I had no backup.

I came down from 92% usage to 60% which is all fine and dandy but I lost the machine. I can't SSH to it now.

I reinstalled network manager but even so, the machine didn't come back up on the network after reboot, so now I'm going to have to reflash the OS on the machine, which is a pain. NVIDIA's support is a joke.

Funnily enough, the machine does boot and it does prompt me to login if I hook up a monitor to it, but for some God forsaken reason it is not possible to use USB keyboards on the Turing Pi V2 for Jetson Nanos due to a limitation. If I had some way to actually login, I could easily fix this problem by re-enabling dhcpd.

[cerebrodigital]

A pesar de la enorme potencia de la supercomputadora, es tan compacta que cabe en la palma de la mano.

Te presentamos sus innovadoras características:

[newsnexttech]

BIOSTAR introduces new AI-NONXS Developer Kit that augments edge AI applications for modern deployment

BIOSTAR has just rolled out its AI-NONXS Developer Kit, a powerful edge AI platform designed for developers and system integrators looking to build and deploy AI-driven solutions at the edge. Supporting NVIDIA Jetson Orin NX and Orin Nano modules, this compact industrial-grade kit is aimed squarely at enabling next-gen AI capabilities in sectors like smart manufacturing, retail, transportation,…

[tannatechbizllp]

Unlocking the Power of NVIDIA Jetson Nano Developer Kit for AI and Robotics | Tanna TechBiz

Leverage the NVIDIA Jetson Nano Developer Kit to build intelligent AI and robotics projects with speed, efficiency, and powerful performance.

[govindhtech]

What Is NanoVLM? Key Features, Components And Architecture

The NanoVLM initiative develops VLMs for NVIDIA Jetson devices, specifically the Orin Nano. These models aim to improve interaction performance by increasing processing speed and decreasing memory usage. Documentation includes supported VLM families, benchmarks, and setup parameters such Jetson device and Jetpack compatibility. Video sequence processing, live streaming analysis, and multimodal chat via online user interfaces or command-line interfaces are also covered.

What's nanoVLM?

NanoVLM is the fastest and easiest repository for training and optimising micro VLMs.

Hugging Face streamlined this teaching method. We want to democratise vision-language model creation via a simple PyTorch framework. Inspired by Andrej Karratha's nanoGPT, NanoVLM prioritises readability, modularity, and transparency without compromising practicality. About 750 lines of code define and train nanoVLM, plus parameter loading and reporting boilerplate.

Architecture and Components

NanoVLM is a modular multimodal architecture with a modality projection mechanism, lightweight language decoder, and vision encoder. The vision encoder uses transformer-based SigLIP-B/16 for dependable photo feature extraction.

Visual backbone translates photos into language model-friendly embeddings.

Textual side uses SmolLM2, an efficient and clear causal decoder-style converter.

Vision-language fusion is controlled by a simple projection layer that aligns picture embeddings into the language model's input space.

Transparent, readable, and easy to change, the integration is suitable for rapid prototyping and instruction.

The effective code structure includes the VLM (~100 lines), Language Decoder (~250 lines), Modality Projection (~50 lines), Vision Backbone (~150 lines), and a basic training loop (~200 lines).

Sizing and Performance

HuggingFaceTB/SmolLM2-135M and SigLIP-B/16-224-85M backbones create 222M nanoVLMs. Version nanoVLM-222M is available.

NanoVLM is compact and easy to use but offers competitive results. The 222M model trained for 6 hours on a single H100 GPU with 1.7M samples from the_cauldron dataset had 35.3% accuracy on the MMStar benchmark. SmolVLM-256M-like performance was achieved with fewer parameters and computing.

NanoVLM is efficient enough for educational institutions or developers using a single workstation.

Key Features and Philosophy

NanoVLM is a simple yet effective VLM introduction.

It enables users test micro VLMs' capabilities by changing settings and parameters.

Transparency helps consumers understand logic and data flow with minimally abstracted and well-defined components. This is ideal for repeatability research and education.

Its modularity and forward compatibility allow users to replace visual encoders, decoders, and projection mechanisms. This provides a framework for multiple investigations.

Get Started and Use

Cloning the repository and establishing the environment lets users start. Despite pip, uv is recommended for package management. Dependencies include torch, numpy, torchvision, pillow, datasets, huggingface-hub, transformers, and wandb.

NanoVLM includes easy methods for loading and storing Hugging Face Hub models. VisionLanguageModel.from_pretrained() can load pretrained weights from Hub repositories like “lusxvr/nanoVLM-222M”.

Pushing trained models to the Hub creates a model card (README.md) and saves weights (model.safetensors) and configuration (config.json). Repositories can be private but are usually public.

Model can load and store models locally.VisionLanguageModel.from_pretrained() and save_pretrained() with local paths.

To test a trained model, generate.py is provided. An example shows how to use an image and “What is this?” to get cat descriptions.

In the Models section of the NVIDIA Jetson AI Lab, “NanoVLM” is included, however the content focusses on using NanoLLM to optimise VLMs like Llava, VILA, and Obsidian for Jetson devices. This means Jetson and other platforms can benefit from nanoVLM's small VLM optimisation techniques.

Training

Train nanoVLM with the train.py script, which uses models/config.py. Logging with WANDB is common in training.

VRAM specs

VRAM needs must be understood throughout training.

A single NVIDIA H100 GPU evaluating the default 222M model shows batch size increases peak VRAM use.

870.53 MB of VRAM is allocated after model loading.

Maximum VRAM used during training is 4.5 GB for batch size 1 and 65 GB for batch size 256.

Before OOM, 512-batch training peaked at 80 GB.

Results indicate that training with a batch size of up to 16 requires at least ~4.5 GB of VRAM, whereas training with a batch size of up to 16 requires roughly 8 GB.

Variations in sequence length or model architecture affect VRAM needs.

To test VRAM requirements on a system and setup, measure_vram.py is provided.

Contributions and Community

NanoVLM welcomes contributions.

Contributions with dependencies like transformers are encouraged, but pure PyTorch implementation is preferred. Deep speed, trainer, and accelerate won't work. Open an issue to discuss new feature ideas. Bug fixes can be submitted using pull requests.

Future research includes data packing, multi-GPU training, multi-image support, image-splitting, and VLMEvalKit integration. Integration into the Hugging Face ecosystem allows use with Transformers, Datasets, and Inference Endpoints.

In summary

NanoVLM is a Hugging Face project that provides a simple, readable, and flexible PyTorch framework for building and testing small VLMs. It is designed for efficient use and education, with training, creation, and Hugging Face ecosystem integration paths.

[elmalo8291]

Elmalo, let's commit to that direction. We'll start with a robust Sensor Fusion Layer Prototype that forms the nervous system of Iron Spine, enabling tangible, live data connectivity from the field into the AI's processing core. Below is a detailed technical blueprint that outlines the approach, components, and future integrability with your Empathic AI Core.

1. Hardware Selection

Edge Devices:

Primary Platform: NVIDIA Jetson AGX Xavier or Nano for on-site processing. Their GPU acceleration is perfect for real-time preprocessing and running early fusion algorithms.

Supplementary Controllers: Raspberry Pi Compute Modules or Arduino-based microcontrollers to gather data from specific sensors when cost or miniaturization is critical.

Sensor Modalities:

Environmental Sensors: Radiation detectors, pressure sensors, temperature/humidity sensors—critical for extreme environments (space, deep sea, underground).

Motion & Optical Sensors: Insect-inspired motion sensors, high-resolution cameras, and inertial measurement units (IMUs) to capture detailed movement and orientation.

Acoustic & RF Sensors: Microphones, sonar, and RF sensors for detecting vibrational, audio, or electromagnetic signals.

2. Software Stack and Data Flow Pipeline

Data Ingestion:

Frameworks: Utilize Apache Kafka or Apache NiFi to build a robust, scalable data pipeline that can handle streaming sensor data in real time.

Protocol: MQTT or LoRaWAN can serve as the communication backbone in environments where connectivity is intermittent or bandwidth-constrained.

Data Preprocessing & Filtering:

Edge Analytics: Develop tailored algorithms that run on your edge devices—leveraging NVIDIA’s TensorRT for accelerated inference—to filter raw inputs and perform preliminary sensor fusion.

Fusion Algorithms: Employ Kalman or Particle Filters to synthesize multiple sensor streams into actionable readings.

Data Abstraction Layer:

API Endpoints: Create modular interfaces that transform fused sensor data into abstracted, standardized feeds for higher-level consumption by the AI core later.

Middleware: Consider microservices that handle data routing, error correction, and redundancy mechanisms to ensure data integrity under harsh conditions.

3. Infrastructure Deployment Map

4. Future Hooks for Empathic AI Core Integration

API-Driven Design: The sensor fusion module will produce standardized, real-time data feeds. These endpoints will act as the bridge to plug in your Empathic AI Core whenever you’re ready to evolve the “soul” of Iron Spine.

Modular Data Abstraction: Build abstraction layers that allow easy mapping of raw sensor data into higher-level representations—ideal for feeding into predictive, decision-making models later.

Feedback Mechanisms: Implement logging and event-based triggers from the sensor fusion system to continuously improve both hardware and AI components based on real-world performance and environmental nuance.

5. Roadmap and Next Steps

Design & Prototype:

Define the hardware specifications for edge devices and sensor modules.

Develop a small-scale sensor hub integrating a few key sensor types (e.g., motion + environmental).

Data Pipeline Setup:

Set up your data ingestion framework (e.g., Apache Kafka cluster).

Prototype and evaluate basic preprocessing and fusion algorithms on your chosen edge device.

Field Testing:

Deploy the prototype in a controlled environment similar to your target extremes (e.g., a pressure chamber, simulated low-gravity environment).

Refine data accuracy and real-time performance based on initial feedback.

Integration Preparation:

Build standardized API interfaces for future connection with the Empathic AI Core.

Document system architecture to ensure a smooth handoff between the hardware-first and AI-core teams.

Elmalo, this blueprint establishes a tangible, modular system that grounds Iron Spine in reality. It not only demonstrates your vision but also builds the foundational “nervous system” that your emergent, empathic AI will later use to perceive and interact with its environment.

Does this detailed roadmap align with your vision? Would you like to dive further into any individual section—perhaps starting with hardware specifications, software configuration, or the integration strategy for the future AI core?

[florinco]

The Nvidia Jetson Orin Nano looks like a computer now.

[zalopro]

Banana Pi BPI-AIM7: powerful, compatible computer Nvidia Jetson Nano

Banana Pi BPI-AIM7 is a new computer (SBC), or more accurate than a calculated module. Those who love to “set the line” want to build a compact system that can take advantage of this calculation module combined with customized board. Powerful performance with RK3588 chip BPI-AIM7 is equipped with ARM RK3588 chipset with four Cortex-A76 cores and four Cortex-A55 cores. The integrated NPU is said…

[makers-muse]

What Are the Must-Have Tools for a Future-Ready STEM Lab in Agartala? 

Introduction: Why Every STEM Lab in Agartala Needs the Right Tools 

A STEM Lab in Agartala is more than just a classroom—it’s a hands-on innovation center where students explore robotics, coding, AI, and engineering. To make learning engaging and future-ready, schools must equip their STEM Lab in Agartala with the right tools and technologies. 

In this guide, we’ll explore the must-have tools that every future-ready STEM Lab in Agartala should have. 

1. Robotics Kits – Powering Hands-On Learning 

A top-quality STEM Lab in Agartala must include robotics kits to teach students about automation, AI, and engineering. Some of the best robotics kits include: 

 LEGO Mindstorms EV3 – Ideal for beginners, offering block-based coding.    Arduino & Raspberry Pi Kits – Great for advanced robotics and IoT projects.    VEX Robotics Kits – Used for competitions and real-world problem-solving. 

These kits help students develop logical thinking and problem-solving skills while preparing them for careers in automation and robotics. 

2. 3D Printers – Bringing Creativity to Life 

A STEM Lab in Agartala should have 3D printers to help students design and prototype real-world objects. Some essential options include: 

Creality Ender 3 – Affordable and beginner-friendly for schools.  Ultimaker 2+ – High-quality prints for advanced projects.  ️ Anycubic Photon – Best for precise, resin-based 3D printing. 

With 3D printing, students can turn their ideas into reality, fostering creativity and innovation. 

3. Coding & AI Learning Kits – Preparing for the Future 

To make a STEM Lab in Agartala future-ready, it must include coding and AI tools for teaching programming skills. Some of the best choices are: 

Scratch & Blockly – Block-based coding for beginners.    Python & Java Programming Platforms – Industry-standard coding languages.  Google AIY & NVIDIA Jetson Nano – AI and machine learning kits for advanced learning. 

These tools help students learn AI, data science, and machine learning, making them ready for future tech careers. 

4. Virtual Reality (VR) & Augmented Reality (AR) – Immersive Learning 

A cutting-edge STEM Lab in Agartala should include VR and AR tools to create immersive educational experiences. The best options are: 

VR and AR tools make learning more engaging and interactive, helping students visualize complex concepts easily. 

5. IoT & Smart Sensors – Learning About the Connected World 

An IoT-enabled STEM Lab in Agartala prepares students for the future of smart technology and automation. Essential IoT tools include: 

Arduino IoT Cloud – Teaches real-world IoT applications.    ESP8266 & ESP32 Microcontrollers – Used for smart device projects.    Smart Sensors (Temperature, Humidity, Motion) – For creating real-time monitoring systems. 

With IoT tools, students can build smart home projects, automated weather stations, and AI-driven devices. 

6. Electronics & Circuit Design Kits – Understanding Engineering Basics 

A future-ready STEM Lab in Agartala must include electronics kits for hands-on engineering projects. The best options are: 

 LittleBits Electronics Kit – Easy-to-use snap circuits for beginners.    Snap Circuits Pro – Teaches circuit design in a fun way.    Breadboards & Multimeters – Essential for real-world electronics projects. 

Electronics kits enhance problem-solving skills and prepare students for engineering careers. 

7. STEM Software & Simulations – Enhancing Digital Learning 

A well-equipped STEM Lab in Agartala should also have digital tools and software for coding, engineering, and simulations. Some must-have software include: 

Tinkercad – Online 3D design and electronics simulation.    MATLAB & Simulink – Used for data analysis and AI applications.  AutoCAD & SolidWorks – Industry-grade design software. 

These digital tools help students practice real-world STEM applications in a virtual environment. 

Conclusion: Build a Future-Ready STEM Lab in Agartala with the Right Tools 

A high-quality STEM Lab in Agartala must include robotics kits, 3D printers, AI and coding tools, IoT kits, VR devices, and circuit design tools to prepare students for technology-driven careers. 

By investing in these essential tools, schools in Agartala can create an engaging, innovative, and future-ready learning environment. 

Want to set up a STEM Lab in Agartala? Contact us today to Upgrade the best solutions for your school!  

[levysoft]

Nvidia ha annunciato il lancio di una nuova versione del suo computer Jetson, denominata Orin Nano Super, venduta al prezzo di 249 dollari, la metà di quello del modello precedente, con l’obiettivo dichiarato di attrarre un pubblico più ampio, composto da hobbisti e piccole aziende.

I computer Jetson sono essenzialmente “cervelli portatili”, progettati per permettere agli sviluppatori di robot, automazione industriale e altre apparecchiature, di eseguire complesse operazioni di intelligenza artificiale direttamente sul dispositivo, senza la necessità di connettersi a un data center remoto.

Una mossa in favore dell’accessibilità

I principali clienti di Nvidia sono di solito grandi compagnie e startup AI, che investono ingenti somme nell’hardware necessario per addestrare e operare modelli di intelligenza artificiale. Con la linea Jetson, Nvidia punta sull’accessibilità, offrendo dispositivi a basso costo a piccole aziende, appassionati e studenti interessati a sviluppare nuovi prodotti con funzioni AI integrate.

In un video promozionale, il fondatore di Nvidia, Jensen Huang, ha modificato il suo consueto stile di presentazione: invece di posare davanti a enormi server che utilizzano chip Nvidia, ha mostrato il nuovo dispositivo Jetson, delle dimensioni di un palmo, presentandolo su un vassoio come se fosse stato appena sfornato.

Caratteristiche del Nuovo Orin Nano Super

Il nuovo Orin Nano Super Developer Kit da 249 dollari raddoppia quasi la velocità e l’efficienza del dispositivo precedente e può elaborare circa il 70% in più di compiti computazionali, secondo Nvidia. Sebbene contenga chip meno avanzati rispetto ai prodotti di fascia alta di Nvidia, è destinato a sviluppatori commerciali che lavorano su tecnologie per consumatori, come droni e fotocamere. Il Jetson Thor di fascia alta, invece, è progettato per supportare robot umanoidi e automazioni sofisticate.

La famiglia NVIDIA Jetson Orin comprende sette moduli con architettura identica, che offrono fino a 275 trilioni di operazioni al secondo (TOPS) e prestazioni 8 volte superiori a quelle dell’ultima generazione per l’inferenza AI multimodale, oltre al supporto di interfacce ad alta velocità.

Il potente stack software comprende modelli di AI pre-addestrati, flussi di lavoro di riferimento per l’AI e framework applicativi verticali, che accelerano lo sviluppo end-to-end per l’AI generativa, nonché per qualsiasi applicazione di AI e robotica edge.

Espansione del portfolio Nvidia

Gli analisti suggeriscono che la linea Jetson potrebbe diversificare l’offerta di Nvidia, specialmente attirando clienti interessati a sviluppare robot. Altre aziende tecnologiche come Intel, Google e Qualcomm offrono sistemi edge simili, che potrebbero competere con il Jetson offrendo soluzioni specifiche per applicazioni come l’elaborazione delle immagini.

Accelera lo sviluppo di applicazioni e la prototipazione con i kit per sviluppatori Jetson Orin

Sono disponibili due kit per sviluppatori Jetson Orin per iniziare a sviluppare e creare prototipi. Il kit per sviluppatori Jetson AGX Orin compatto offre le massime prestazioni e può anche emulare qualsiasi modulo Jetson Orin. Il kit per sviluppatori Jetson Orin Nano è ancora più piccolo e include una scheda carrier di riferimento compatibile con tutti i moduli Orin NX e Orin Nano.

Intervistato sui concorrenti, Deepu Talla, vicepresidente di robotica e edge computing di Nvidia, ha affermato che il prodotto è pensato per scopi generali e può eseguire “tutti i più recenti modelli generativi di AI”.

Disponibilità in Cina

Nonostante le restrizioni imposte dagli Stati Uniti alla vendita dei componenti più avanzati di Nvidia in Cina, l’azienda ha dichiarato che il nuovo prodotto Jetson sarà disponibile nel paese attraverso distributori locali.

[dailyengenharia]

PESSOAS COMENTARAM SOBRE O MERCADO ATUAL [ENGENHARIA / DEVS] - DICAS E ANÁLISES

MERCADO ATUAL

As maiores demandas são para escrita, fazer cursos especializados, além de entender o funcionamento de IA Generativa.

Como Junior Data Scientist

Possua conhecimento em Estatística e Probabilidade;

Possua conhecimento em ferramentas de visualização de dados e/ou Business Inteligence com capacidade em exploração e pré-visualização de dados;

Tenha conhecimentos básicos em algumas Linguagens utilizadas na área de data science e do processo de experimentação e avaliação de modelos de Machine Learning;

Possua conhecimentos em SQL e Banco de Dados Relacionais;

Possua conhecimentos em :

Visão computacional

Aprendizagem de máquina com foco em processamento de imagens

SLAM

Python / C++

Algoritmos de localização e mapeamento

Filtro de Kalman

Fusão de dados (fusão de tipos de dados distintos)

É legal que você...

Tenha inglês intermediário leitura, escrita, escuta e fala.

Tenha conhecimento em:

Conhecimento de ROS

IA embarcada (Ex.: Jetson Nano NVIDIA)

Sistemas embarcados/robótica

—-----------------------

2. Estude continuamente os temas cobrados. Leia o que se pede na vaga e estude cada ponto. “Ah mas não dá tempo para ver tudo “. Ok, mas também não dá para ficar parado. Existem assuntos que sempre cobram como CI/CD, deploy, monitoramento, tradeoff de viés e variância, retreino, concept drift, data drift, docker, docker compose, como servir e criar api, streaming, batch, pyspark, etc. estude um por vez.

3. Tenha seus padrões de código básicos, uma pipeline padronizada, uma configuração básica de ambiente, poetry.

4. Estão pedindo muito llm. Sugiro rever os conceitos de rag, fine tuning, sLM, se você já souber os fundamentos muito bem.

5. Faça leet code e hacker rank. Alguns falam “ah, mas não usa no negócio”, ok, mas vai cair no teste. Quer passar ou arrumar motivo para não estudar?

6. Peça dicas a quem já passou por isso. Eu pedi ao Ricardo Costa e ele me deu todas essas dicas e mais um pouco. 

7. Acompanhe canais com live code na area. Daniel Bourke , Sebastian Raschka, PhD , Andrej Karpathy, Téo Calvo no Téo Me Why , Hallison Paz, PhD , Kizzy Terra , Programação Dinâmica e vai codarrrrr.

8. Seja fluente em scikit-learn, pandas, (pyspark e PyTorch é minha sugestão mas tem gente que não concorda. Se for difícil usar pyspark em seu pc, use polars)

9. Modularize seus códigos e crie pipelines com metaflow ou com programação funcional mesmo.

10. Treine inglês diariamente. Sugiro contratar um professor nativo. Clarke - Professor de inglês é um excelente professor e com ele eu aprendi demais.

[bird4669]

Introducing NVIDIA Jetson Orin™ Nano Super: The World’s Most Affordable ...

[elmalo8291]

Elmalo, let's move forward with scoping a full pilot buildout—starting with the v1 Mars Habitat Monitor. This path offers a compelling, high-stakes testbed for the Iron Spine system and allows us to prototype under extreme, failure-intolerant conditions. Designing for Mars pushes the architecture to its limits, ensuring resilience, autonomy, and layered intelligence from the outset.

🚀 v1 Mars Habitat Monitor – Pilot Buildout

🔧 Environmental Design Requirements

Radiation-Hardened Components: Select radiation-tolerant MCU/FPGA and sensor components (e.g., RAD750 derivatives or Microsemi FPGAs).

Thermal Regulation: Passive and active methods (phase-change materials, aerogels, thin-film heaters).

Dust Protection: Hermetically sealed enclosures with electrostatic or vibrational dust mitigation (similar to the Mars 2020 rover’s approach).

Power Constraints: Solar panels + supercapacitors for charge buffering, with ultra-low power idle modes.

Communications Delay Tolerance: Incorporate DTN (Delay-Tolerant Networking) bundles for relayed Earth-Mars messaging.

🧠 Sensor Suite

Life Support Monitoring:

CO₂ / O₂ / CH₄ levels

Humidity / Temperature / Pressure

Structural Integrity:

Microfracture sensors (piezo-acoustic or fiber optic strain gauges)

Vibration analysis (accelerometers/IMUs)

Radiation Exposure:

Ionizing radiation detectors (Geiger-Müller tubes or RADFETs)

Environmental:

Dust density (LIDAR or IR scattering)

UV exposure, ambient EM fields

🧩 System Architecture

Sensor Synchronization:

Use local PTP clocks with oscillator drift correction from a central unit

Redundant clocks for fault detection

Data Fusion Layer:

Edge-level Kalman filters for each sensor type

Onboard anomaly detection with TensorRT-lite or ONNX models for pattern deviation

Edge Inference:

PyTorch Mobile or ONNX on an ARM-based SOM (e.g., NVIDIA Jetson Nano with radiation shielding)

Data Relay and Storage:

Store all telemetry in redundant onboard buffers

Use prepackaged DTN bundles and error-correcting transmission protocols to transmit during windowed comms

Digital Twin Integration:

Each monitor syncs with a simulated Mars habitat replica in Earth orbit or ground control

Dashboard via Prometheus/Grafana or a bespoke Unity/Unreal-powered 3D monitoring console

⚙️ Resilience & Power

Watchdog Timers: Auto-reset on failure, with a secondary power rail for minimal heartbeat signal

Power Monitoring: Real-time tracking of draw, supercapacitor status, and solar charge

Failover States: Downgraded functionality tier for power preservation and sensor core redundancy

Would you like a modular breakdown of the above, suitable for PCB design, mechanical housing, and software stack implementation? Or perhaps a visual mission profile that tracks expected operational phases, anomaly response, and maintenance cycles for the Mars system?

Let’s make this pilot the template for Iron Spine’s interplanetary leap.

[hackernewsrobot]

Nvidia Jetson Orin Nano Super [video]

https://www.youtube.com/watch?v=S9L2WGf1KrM