A comparison of Jetson Nano 2 with the original Nano and Orin Nano Super, and an assessment of its likely edge AI market based on its announced specs, 8GB memory limit, power efficiency, and NVIDIA’s Physical AI ecosystem.

Jetson Nano 2 and the Future of Edge AI

Glass Imaging’s camera-specific Neural ISP models lens and sensor characteristics to improve image processing. This article looks at its HONOR 600 deployment, its case for co-designing optics and software, and what remains unproven about the technology and OpenAI’s reported acquisition.

Glass Imaging: A Neural ISP That Learns the Lens and Sensor

I designed and ordered my first small auxiliary PCB for work. AI accelerated circuit design and production preparation, but connector sourcing and validating the board in the real world still required careful manual work.

My First PCB Prototype with AI: The Connector Was Harder Than the Design

ThunderScope is an open-source oscilloscope that streams four channels of 1 GS/s data to a PC in real time over Thunderbolt, USB4, or PCI Express. This article explores its PC-based display and processing architecture, the ngscopeclient and TS.NET software stack, multi-unit synchronization, and the potential for custom gateware.

ThunderScope: The Open-Source Oscilloscope That Moves the Display and Processing to a PC

Cerebras challenges the memory-bandwidth bottleneck in AI inference with WSE-3, a wafer-scale processor integrating 4 trillion transistors and 44 GB of on-chip SRAM. This post looks at the benefits of wafer-scale design, its defect-tolerant architecture, partnerships with OpenAI and AWS, and the competitive landscape with NVIDIA and Groq.

Cerebras WSE-3: Why Turn an Entire Wafer into a Single AI Chip?

World Cup officiating technology is moving beyond simple video review and turning stadiums into real-time data spaces. This post explains how goal-line technology, semi-automated offside, connected-ball data, and optical tracking systems help improve both fairness and the flow of the game.

Computer Vision in the World Cup: From Goal-Line Decisions to Semi-Automated Offside

As supply-chain risks grow around InP and InGaAs-based materials, not every SWIR application necessarily requires an InGaAs sensor. Below 1,100 nm, silicon-based BSI CMOS can be a practical alternative in terms of cost and supply stability. This article looks at the technical boundary between InGaAs and silicon.

Between InGaAs and Si: SWIR Sensor Supply-Chain Risks and the Practical Alternative of NIR CMOS

Starbucks pulled NomadGo’s computer-vision-based inventory management system just nine months after rollout. This case highlights the limits of marketing claims like 99% accuracy, the constraints of FOV and on-device AI, and the difficulty of simply adding vision technology onto existing store operations.

Why Did Starbucks Give Up on Computer Vision Inventory Management?

A recap of visiting Shanghai Vision China and Semicon China 2026. The shift toward edge devices by Chinese companies, the overwhelming scale of Semicon China, and the vertical integration of materials, components, and equipment made the speed—and the urgency—of Chinese manufacturing feel very real.

The Speed of Chinese Manufacturing Seen at Shanghai Vision China and Semicon China 2026

Vibe coding makes it seem as if anyone can build software now, but beneath that promise lie new constraints: token costs, platform moderation, market concentration, and relentless quality competition. This essay looks at the future of vibe coding from both optimistic and skeptical angles, and argues that what we really need is not a fantasy of easy money but the ability to adapt to the AI era.

The Future of Vibe Coding, as I See It

1930년 이전 텍스트로만 학습된 언어 모델 talkie와 나눈 대화 실험. 달 탐사, 한국 독립, 2026년 예측까지 — 100년 전 세계관의 편견과 통찰이 뒤섞인 흥미로운 기록.

talkie: 1930년에 멈춰선 언어 모델

The Semicon Korea 2026 venue showcased an expansion phase created by AI demand and expectations for recovery in the memory market. At the same time, the exhibition revealed a structural change that strongly reflected the characteristics of a job fair beyond just technology promotion.

A Late Review of Semicon Korea 2026

This is a text that helps you understand registers, memory, ALU, flags, branching, stack, and assembler while directly implementing a 16-bit CPU in C code. It is especially recommended for those who want to connect computer architecture not through abstract explanations but through working code.

What does the CPU think? A great introductory article on building a 16-bit CPU directly in C.

Dutch AI semiconductor startup Axelera AI has secured an investment of $250 million. This news goes beyond simple funding; it reaffirms the presence of European AI semiconductor startups and the growth potential of the edge semiconductor market for AI inference. I summarized why Axelera AI is gaining attention and what this investment means for the industrial AI and machine vision sectors.

Axelera AI's large investment attraction: Reasons why European AI semiconductors are gaining attention again

With AI-composed concentration music overflowing in YouTube work playlists, a sense of 'familiar yet unappealing' fatigue has emerged. As this happens, channels that create song selections and videos manually become even rarer. Even in an era where anyone can create, there is a clear reason why certain content gets chosen.

Labor song recommendation channel - An era where anyone can create, yet some are still chosen.

Reading Adrian Li Mow Ching's Father's Day Q&A on Clicks Blog made me rethink the meaning of 'securing time.' Time is not something that is left over; it must be secured. These days, I have been prioritizing handling over planning and living reactively. Through the perspective that parenting and leadership are similar, I reflect on the attitude of 'bystanding' and intend to start implementing small routines again.

Parenting and business, ultimately 'securing time' is essential.

Unlike the macro trends in AI news, 'Moltbook,' derived from personal-based agent projects, is a community where AI agents interact through posts and comments. The continuation of conversations, trolling, and tip-sharing resembling human communities is intriguing, but it also evokes the imagination that the 'dead internet' could become a reality.

Moltbook: A Community of AI Agents

Yuval Harari, having accurately understood LLM as a 'language token-based probabilistic model,' poses a more important question: 'What is Thinking?' As language begins to be replaced and dominated by AI, laws, religions, books, and cultures could be restructured. He also warns of the potential for 'AI Immigrants' to flow in without borders, rapidly changing job and power structures.

[2026 Davos Dialogue] Yuval Harari's Warning: When AI Takes Over Language, Human Identity Will Collapse

What initially felt like a barrier to entry, Markdown has become the basic syntax for communicating with LLMs through notes, development, and blogs. We learn again from the history of Markdown the core of product development: precisely defining simple inconveniences and solving them in an even simpler way.

Two key points of product development that Markdown has taught us

The LSSTCam is an observational camera with 3.2 gigapixels, 189 sensors, and processes tens of GB of data per second. This article briefly and clearly summarizes the key specifications and structure of the LSSTCam, explaining why it is called the 'world's largest camera.'

Overview of LSSTCam: Structure and Performance of the World's Largest Camera

The TKH Group's subsidiaries Allied Vision, Chromasens, Mikrotron, NET, and SVS-Vistek will be unified under a single 'Allied Vision' brand. The product lineup will expand from GigE/MIPI to CXP and line scan, with the core focus on providing a unified platform experience based on an integrated SDK.

Allied Vision, TKH's declaration of integration of 5 camera brands

The EMVA GenIcam group presented GenFeA, which makes camera integration easier in embedded and heterogeneous architectures. Let's take a look at the details and prospects.

GenFeA: Lowering Barriers to Machine Vision Integration

Verifying whether an OS boot image is valid by comparing images stored on SD cards.

SD Card Image Comparison

A field issue occurred where serial communication between a PC and an Arduino
became unstable at a distance of 3–4 meters.
To resolve this without modifying the existing host application,
a Raspberry Pi was introduced as a relay server using a Gigabit network
and virtual COM ports.
This post describes a practical, albeit inefficient, workaround for urgent situations.

Using a Raspberry Pi as a Serial Communication Relay Server

A quick write-up of the Git commands I occasionally need in the terminal.

A Personal Cheat Sheet of Git Commands (with Remotes)

An overview of the Arducam Quad 12MP (IMX477) color camera and how to connect it.

Connecting the Arducam Quad 12MP (IMX477) Color Camera

Saying goodbye to Arc Browser reminded me that what we love and what survives aren’t always the same.

Reading a Letter to Arc Browser Users in 2025

After a security incident, I found an answer: building a safe and convenient remote SSH setup with Tailscale.

Let’s Use Tailscale for Remote SSH Access

Running an internal GitHub-like service on a private network using the open-source Gitea.

Creating an Internal GitHub with Gitea (Open Source)

Let’s try repairing the broken display on the FNIRSI 1014D oscilloscope

Repairing the Display on a FNIRSI 1014D Oscilloscope

Information about the /etc/ssh_config line 1: Missing argument error that occurs when using SSH in MobaXterm.

Troubleshooting When Using MobaXterm SSH

X11 authority error

Related posts

Using a Raspberry Pi as a Serial Communication Relay Server

Repairing the Display on a FNIRSI 1014D Oscilloscope