LodeHQSubscribe →

TypeScript goes native, Chrome gets ARM64 Linux

Dev · 2026-07-27

Tools & Platforms
Interactive PGSimCity Visualizes PostgreSQL Engine Internals for Deeper Insight1 MIN

PGSimCity is a free, interactive web visualizer that walks a SQL query through PostgreSQL’s planner, executor and storage layers. It lets developers and DBAs see exactly where data moves, making the engine’s inner workings concrete and easier to debug. The prototype is community‑maintained, so contributions improve its accuracy.

Scriptc lets TypeScript run as a zero‑runtime native binary4 MIN

Vercel's Scriptc compiles plain TypeScript into tiny native binaries that contain no JavaScript engine, cutting startup time to milliseconds. It works without code changes, statically compiling up to 99% of statements, and falls back to an embedded QuickJS engine only for truly dynamic parts.

Google Chrome Gains Official ARM64 Linux Build with Native Widevine3 MIN

Google finally ships an official Chrome v150 for ARM64 Linux, including native Widevine DRM. The build installs via apt on Ubuntu 26.04, brings full Google sync and DRM playback to Raspberry Pi 5 and other ARM laptops, closing the feature gap that forced users to rely on Chromium work‑arounds.

LiteRT.js Brings Native‑Speed AI Inference to the Browser4 MIN

Google released LiteRT.js, a WebAssembly‑based JavaScript binding of its LiteRT runtime that runs .tflite models in any browser with hardware acceleration. It outperforms TensorFlow.js, giving developers ultra‑low latency, client‑side privacy and no server costs for tasks like text generation or object detection.

Embed Super Mario 64 physics into any engine with libsm641 MIN

libsm64 packages reversed Super Mario 64 movement and rendering code into a shared library that drops straight into Unity, Godot, Rust, Python, and more. Developers can embed Mario with minimal effort, only needing the original ROM for textures and animations. It opens up classic platforming physics for any custom engine or mod.

AI-Assisted Development
Genkit’s Agents API collapses chatbot plumbing into one reusable interface7 MIN

Google’s open-source Genkit now bundles conversational state, tool loops, and streaming into a single Agents API. Developers can define an agent once and run it on server or client with built‑in persistence, cutting boilerplate and accelerating multi‑turn AI app builds.

Google adds auto‑evaluation loop to keep coding agents from silently breaking10 MIN

Google’s new coding‑agent skill automates a five‑stage evaluation loop, data prep, run, analyze, fix, verify, so prompt tweaks can’t silently degrade your agents. By embedding AutoRaters and independent scoring, it forces disciplined, metric‑driven improvement, catching regressions before they hit production.

Treat AI code generators as subsystem maintainers, a Linus Torvalds playbook6 MIN

The post argues that true automatic programming requires developers to assume Linus Torvalds' leadership role while AI agents act as maintainers of individual subsystems. This turns AI from a copy‑paste tool into an orchestrator of complex kernel work, accelerating progress but still demanding human oversight for overall coherence.

Kimi K3: 2.8 T‑parameter Mixture‑of‑Experts model with 1 M‑token context and native vision137 MIN

MoonshotAI’s Kimi K3 pushes open LLMs to 2.8 trillion parameters, activating up to 104 billion weights per token via a 896‑expert MoE. It adds a 1‑million‑token context window and built‑in vision, delivering frontier‑level coding, reasoning, and agentic performance while remaining open‑source.

Engineering Practice
Prefactoring lets you clear code roadblocks before building new features1 MIN

Prefactoring, a pre‑emptive refactor, reshapes code before a new feature is added. By separating cleanup from logic, teams cut review time, reduce bugs, and keep rollbacks simple, letting them ship faster and safer.

Misago slashes frontend bloat by swapping React for htmx3 MIN

Misago replaced its duplicated React UI with htmx, shrinking the codebase and cutting client‑side payloads. By turning server‑rendered Django templates into dynamic islands, the forum regained fast loads on older devices and simplified plugin development. The switch also eliminated duplicate translation files and a heavy JavaScript build step.

Canva’s in‑memory session revocation system scales to hundreds of millions10 MIN

Canva stores encrypted session data in browser cookies and keeps a 12‑hour revocation list in memory, avoiding per‑request DB lookups. To stop deployment‑time cache stampedes, they switched from MySQL seeding to an S3‑backed cache, slashing startup latency while preserving sub‑millisecond revocation checks.

Get Dev in your inbox, every issue.
Subscribe free
Get the app · Privacy · Terms · About · Contact
© 2026 LodeHQ