Page 3. Posts are ordered by date, with each page loading a bounded set of covers.
In a WSL environment with Ubuntu 26.04 LTS and an RTX 4060 Laptop GPU, this note uses micromamba to prepare CUDA 12.4, GCC 13, and cuBLAS without modifying system directories, builds the llama.cpp CUDA backend, and compares CPU and CUDA speed with Qwen3 4B.
On a local development machine with an RTX 4060 Laptop GPU, 8GB VRAM, and 32GB memory assigned to WSL, I tested Qwen3 4B, Qwen3 8B, Gemma 4 E4B, and Gemma 4 12B for load time, generation speed, VRAM usage, multimodal capability, and practical trade-offs.
The real value of local open models is not a benchmark score, but whether they can enter tasks that happen every day. This article puts Qwen3 4B, Qwen3 8B, Gemma 4 E4B, and Gemma 4 12B into three workflows: development assistance, image understanding, and writing organization.
A continuously updated local open-model installation note. Starting from a Windows + Ubuntu 26.04 LTS development machine with 64GB host memory, 32GB assigned to WSL, and an RTX 4060 Laptop GPU with 8GB VRAM, this article first decides whether Ollama should run on Windows or WSL, compares llama.cpp, LM Studio, vLLM, and other options, and then lists models worth keeping locally.
Starting from Andrew Ng's comments on small teams and high-context engineers, this article explains why AI pushes senior developers toward a broader product full-stack: problem framing, prototype, code, copy, compliance, launch, operations, and iteration.
The second essay in Thirty Years in IT and Me. In 1997, I wrote a small JavaScript program that made a butterfly fly across a web page. It was tiny, but it flew through the browser war, the rise of DHTML, and the beginning of ECMAScript standardization.
Starting from a Windows development machine with an NVIDIA GPU, this article explains how to verify CUDA in WSL, run a small PyTorch test, place local model caches, and debug cases where the GPU is not actually being used.
A practical migration note for a mixed Python and Node project: move the repository from a Windows partition into the WSL Linux filesystem, then redraw the boundaries around files, paths, process calls, and performance.
Starting from a nearly full C drive, this article explains why WSL2 ext4.vhdx files grow but do not immediately shrink, how to locate Docker, node_modules, model caches, and package caches, and how to clean and compact them safely.
Using a local API service as the example, this article explains Docker Desktop's WSL backend, bind mounts, named volumes, hot reload, logs, and disk-space hygiene.
Starting from a team development-environment migration, this article explains the boundaries of WSL1, WSL2, distributions, VHDX storage, Windows integration, systemd, networking, and filesystems, so you can decide what belongs in WSL.
Using a small image batch-processing utility as the thread, this article explains Qt, GTK, and OpenCV GUI development under WSLg, including environment checks, debugging, file boundaries, packaging choices, and team usage.
Using a real log investigation and batch file-cleanup workflow, this article explains how to connect rg, jq, awk, fzf, make, timers, editors, and Windows tools into a stable WSL-based workflow.
Starting from xeyes, gedit, and remote windows, this article explains the roles of X11, Wayland, XWayland, Weston, and RDP inside WSLg, and why old DISPLAY-based tutorials can now be counterproductive.
The first essay in Thirty Years in IT and Me. From a 386, PDP-11, Slackware Linux, and ray tracing coursework to assembly-level development, the browser war, and the Web before standards settled down.
This is not about secret indicators. It is a practical knowledge map for programmers who want to build a quant-trading platform: data, factors, strategies, backtesting, risk controls, execution, performance, quality, and the long path from research to real results.
Starting from a simple fim image.png command, this article explains why WSLg can display Linux GUI programs directly on the Windows desktop, and how that helps with image previews, computer-vision work, and small X11 tools.
A practical test of gemma4:12b audio input on a local test machine: the model metadata declares audio support, but Ollama's native /api/chat path could not reliably accept audio; the working path was the OpenAI-compatible input_audio block.