← digests › te9.dev
[ digest / last-10 ]

Last 10 Analyzed.

The 10 most recently analyzed bookmarks from the te9.dev archive. Each entry has been crawled, parsed, and annotated by an LLM for relevance, purpose, and practical use.


Nanolathe

purpose

Nanolathe is an open-source, MIT-licensed 2.5D real-time strategy engine that clean-room reimplements the classic Total Annihilation, playable directly in a browser or natively on Windows, Mac and Linux. It offers a modern GPU renderer, classic renderer mode, online multiplayer with friends, and published research on the original game's formats.

when to use

It is most valuable when studying or building browser-based game engines, GPU rendering techniques, or real-time multiplayer architecture. It is also useful for developers wanting an open codebase to mod, extend, or use as a model for clean-room reverse engineering projects.

tags
game engine RTS open source MIT license GPU rendering browser game multiplayer game development Total Annihilation

Agent Plugins

purpose

Defines an open, vendor-neutral specification (v1.0.0) for packaging AI agent extensions as portable plugins, standardizing a directory structure with a plugin. manifest, an Agent Skills folder, and an mcp. descriptor for MCP servers.

when to use

Use this when creating reusable AI agent components (skills, tools, or MCP servers) that must interoperate across multiple clients like Claude, Cursor, or Copilot, rather than targeting a single vendor's proprietary plugin format.

tags
AI agents open standard specification MCP Model Context Protocol Agent Skills plugin format interoperability developer tooling packaging

Ensu

purpose

Ensu is a private, personal LLM application from Ente that runs locally on your device and is designed to grow and adapt with you over time. It is available across desktop platforms (macOS, Windows, Linux) and mobile platforms (Android, iOS).

when to use

It is most valuable when you need AI assistance while handling sensitive or private data, working offline, or wanting to avoid cloud-based AI services. It is also useful for developers exploring local LLM implementations and on-device AI patterns.

tags
LLM AI assistant local-first privacy on-device AI personal assistant offline desktop app mobile app Ente

nitzangames/SunsetCity: Sunset City: the original Fable 5.1 High-effort one-shot browser driving sandbox, with source, tests, prompt and build history.

purpose

It is an open-source repository containing the full source, tests, build history, and original prompt for 'Sunset City', a browser-based open-world driving sandbox set in a pastel low-poly city at golden hour. The project includes 92 logic tests, headless Chrome screenshot captures, and a reviewer API (window.__sim) for scripted, reproducible game inspection.

when to use

Most valuable when building or architecting a browser game or simulation and needing a reference for testable, modular ES-module code with an esbuild pipeline. It is also useful when setting up automated headless-browser verification (deterministic sim, screenshot assertions, console error checks) for interactive web apps.

tags
three.js browser game open source esbuild game development headless testing ES modules deterministic simulation AI-generated code driving sandbox

Agent Standup — your AI agents in a stand-up call

purpose

Turns Claude Code sessions into a live, simulated stand-up meeting where a lead agent gives orders and sub-agents join, share their screen, and report back. It displays per-agent status, context gauges, token usage, replays, recaps, and an agent tree, plus voice and custom character features.

when to use

Most valuable when orchestrating multiple Claude Code sub-agents in parallel and you need an at-a-glance view of agent statuses, dependencies, and blockers. Also useful for reviewing past sessions via replay, recap, and the agent tree.

tags
AI agents Claude Code agent monitoring developer tools visualization VS Code extension open source productivity multi-agent orchestration

lxfater/inpaint-web

purpose

It is a free, open-source browser-based tool for image inpainting (removing unwanted objects from photos) and image upscaling (super-resolution), powered by WebGPU and WebAssembly with no server-side processing required.

when to use

It is most valuable when you need to remove unwanted elements from images or enhance image resolution without uploading files to a server, or when researching how to implement WebGPU/WASM-powered ML inference in web applications.

tags
inpainting image editing webgpu wasm machine learning super-resolution image upscaling open source browser-based client-side ML

microsoft/BitNet

purpose

BitNet (bitnet.cpp) is Microsoft's official open-source inference framework for 1-bit and 1.58-bit LLMs, providing optimized kernels that deliver up to 6.17x speedups and up to 82.2% energy reduction on CPUs compared to full-precision models, capable of running a 100B parameter model on a single CPU.

when to use

Use this when you need to deploy LLM or embedding capabilities on CPUs, edge devices, or energy-sensitive environments where GPU resources are unavailable or too costly. It is especially valuable for local-first applications, prototyping, or scenarios requiring 1-bit model efficiency without sacrificing output quality.

tags
LLM AI inference machine learning quantization 1-bit models edge computing CPU optimization open source Microsoft local AI

crow-hq/llm-wiki-v2

purpose

It ingests .txt, .md, and .pdf documents, automatically files them into an organized wiki of Markdown notes, and answers questions with clickable citations. It uses a CROW typed-classifier to keep ingestion fast and cheap, and works with any LLM provider including free local models via Ollama.

when to use

Use it when you need a self-hosted, privacy-friendly alternative to cloud RAG services, such as building internal documentation search, meeting-note Q&A, or knowledge assistants on top of your own documents. It is especially valuable when you want low token costs, local model support, or citations pointing back to source notes.

tags
LLM knowledge-base RAG self-hosted Python HTTP API markdown docker local-first open-source

Discretionary Review

purpose

Discretionary Review is a city-building simulation game by Oliver Sin in which players attempt to navigate San Francisco's housing approval process to get housing built.

when to use

It is most useful during downtime for fun, or when brainstorming approaches to gamification, interactive narrative design, or civic engagement web projects.

tags
browser game simulation city-building housing policy San Francisco civic tech interactive storytelling gamification

wallhop

purpose

WallHop is a free paywall bypass tool that lets users read paywalled news articles by simply pasting a link, with no signup required. It offers multiple access methods including URL prefixing, a browser bookmarklet, an iPhone Shortcut, and API/raw endpoints.

when to use

This is most valuable when you need to read a one-off paywalled article someone shared, or when building tools that need to retrieve article text for research, archiving, or content analysis purposes.

tags
paywall bypass article reader API bookmarklet content access news web scraping reading tools iOS shortcut free tool