• My Keyboard and Mouse Vanished in Windows, While Linux Was Fine

    In Windows the keyboard and mouse were completely dead — no cursor, no keystrokes. Rebooting into Linux they worked fine. Mounting the NTFS partition read-only from Linux and parsing the registry hives and event logs led to a dangling LowerFilters entry under HIDClass, left behind when I deleted Steam's files by hand during a migration; cleared with reg commands from WinRE.

  • Steam Does Nothing When You Click It: Tracing Two Layered Faults

    apt installed Steam, but clicking it did nothing. The logs showed a severed download path and missing 32-bit libraries — and tracing the second fault led to a shadowed python3 on PATH that made steamdeps fail silently.

  • Tree-Walking Interpreter Node Architecture (Part 5): The Boundary of the Node Contract — Contexts, Failures, and Environment Mirroring

    The Node contract guarantees only positions and rendering; dispatch lives on Expr / Stmt methods. This part stands at the contract boundary: first the host and environment mirroring (why the two-environment mirror is the cost of the pre-check), then the contract itself (immutability and invariants), and finally the two things outside the interface.

  • Tree-Walking Interpreter Node Architecture (Part 4): Order and Depth — Descent Position and Recursion Bound

    Preorder and postorder do not come from traversal parameters; they come from whether the method body writes itself first or takes the child results first. A left-associative chain never touches the stack; recursion remains only in right subtrees, parentheses, and unary operators, so the limit sits only at the parser.

  • Tree-Walking Interpreter Node Architecture (Part 3): Unexported Methods and Two Extensions — Who Guarantees Exhaustiveness

    Three interfaces, six methods, only three exported; unexported methods keep the shape set extensible only inside the package. Adding a read versus adding a shape: who guarantees exhaustiveness in each case? Finally, five mainstream designs put the choice to the test.

  • Tree-Walking Interpreter Node Architecture (Part 2): Positions and Errors — How Coordinates Enter the Interface

    A position is not another operation but a projection on the interface; why it must be a method rather than a field, why a single point rather than a range, and why the layer name is left to the host.

  • Tree-Walking Interpreter Node Architecture (Part 1): Three Reads — All Implemented as Node Methods

    Why evaluation, checking, and rendering all live as node methods; why Node / Expr / Stmt split into three interfaces; what each layer rejected and where its cost lands. Positions and openness are left to the next two parts.

  • Content Only Shows Up After a Refresh: A CORS Postmortem in Three Layers of Disguise

    Client-side navigation to an article rendered an empty skeleton, while pressing F5 showed everything. The backend CORS allowlist pinned `http://localhost:3000` while the browser was on `http://127.0.0.1:3000`. A postmortem of why a refresh helps, why the app layer stayed completely silent, and how to prevent it on both the Gin and the useAsyncData side.

  • A Nuxt.js SSR Memory Leak Postmortem: How One Global Variable Crippled a 2GB Server

    From a cloud disk alert and unresponsive SSH, this postmortem digs through Swap, containers, and the Node process to find the real culprit: a module-level global state leak in a Nuxt.js SSR app—plus the fix and verification.

  • Memory: Multi-Session Facts and Lightweight RAG (Part 5)

    Giving the agent two kinds of memory: cross-session key-value facts (remember/recall, persisted to a JSON file) and lightweight RAG over local notes (embed → cosine → topK), with measurements showing why a fixed similarity threshold and the embedding prefix are the two most over-trusted details.

  • Per page
    itemsFound 76 documents