Knowledge Share
Technical articles, tutorials, and insights
LLM-Driven Automated Testing Series (14): Production Engineering — Why Testing Turns Out to Be the Cheapest Line Item in This Whole Series
A code-generation workflow I documented earlier burned thousands of dollars in tokens during debugging. The testing scenarios verified across this series — AutoRestTest testing a 15-operation API for about a dime, pytest-triage capping itself at 10 model calls per run by default — are off by orders of magnitude. This article puts the evidence from 13 prior articles side by side to answer the three questions the planning doc left for last: why testing-scenario cost is what it is, where the human-in-the-loop gate actually belongs, and which scenarios should explicitly avoid LLMs altogether.
LLM-Driven Automated Testing Series (13): Flaky Tests Aren't a Rerun-Count Problem, They're a Classification Problem
pytest-rerunfailures (477 stars) and flaky (397 stars) handle test instability with the same one-liner: if it fails, rerun it a few times. That doesn't answer what the planning doc actually asks — was this failure a real bug, environment jitter, or a badly designed test? The only project actually attempting that root-cause classification with an LLM is a 2-star, less-than-two-months-old repo called pytest-triage — and its most instructive feature turns out to be the four safety invariants it wraps around AI judgment, guaranteeing AI is never allowed to affect whether a test actually passes or fails.
LLM-Driven Automated Testing Series (12): Beyond Pixel Diffs — Who's Actually Making LLMs 'Understand' UI Changes
BackstopJS, at 7.2k stars, handles rendering noise with exactly four numeric config knobs — misMatchThreshold, requireSameDimensions, ignoreAntialiasing, usePreciseMatching — which is still just fighting font-rendering jitter with pixel-level tolerance. The project that actually wires a VLM/LLM into the visual-regression judgment loop is a 23-star repo called vlmkit, whose own internal benchmark notes state outright that one model hallucinates red into red. Putting these two side by side answers, concretely, what semantic visual regression actually solves — and what new problems it introduces.
One Open Source Project a Day (No. 256): Anatomy — Have Claude Invent a Machine for Any Technical Concept and Draw It as an Interactive Isometric Illustration
Anatomy is an Agent Skill. Give it a technical concept and it invents a physical device whose behavior embodies that concept (a cache key becomes a pin-tumbler lock, a shared GPU context becomes a gantry plotter), then drives the picture from a small model to produce an interactive isometric SVG illustration, with optional WebGL shaders and 3D. A dependency-free drawing kit, a nine-step workflow, and built-in geometry and overlap audits. Works with Claude Code, Codex, OpenCode, and more. MIT License, about 25k Stars.
One Open Source Project a Day (No. 255): Mika — An AI Agent for Android That Operates Your Apps and the Web on a Hidden Background Screen
Mika is an open-source Android AI assistant. You give it one instruction, and it looks at the screen, finds buttons, taps, and types like a person across apps and the web, on a hidden background display so your main screen stays free. There is no Mika server; requests go only to the model provider you configure. It ships presets for DeepSeek, Qwen, Kimi, Claude, Gemini, OpenAI, Ollama and more, plus a built-in browser, scheduled tasks, a Linux sandbox, and MCP/CLI access. Written in Kotlin. Apache-2.0 License, about 7.2k Stars.
One Open Source Project a Day (No. 257): AI Design Skills — Design Guidelines Written as Rules an AI Can Execute, for Landing Pages and Figma
AI Design Skills is a set of design Skills from elayadesign. Each Skill is a self-contained folder, and there are two so far - landing-page-design (the full landing page flow from strategy and copy to visual system) and figma-designer (structured Figma work using variables, text styles, auto layout, and component states). The rules go down to font allowlists, spacing steps, easing curves, and banned words. Works with Claude Code, Cursor, Codex, Windsurf, and more. MIT License, about 14k Stars.
One Open Source Project a Day (No. 254): NVIDIA Auto Ontology — A Certified Semantic Layer for Data Agents That Says 'I Can't Answer That' When It Can't
Auto Ontology is NVIDIA's open-source Python framework that builds an ontology of terms, SQL attributes, and table relationships from enterprise metadata, stores it in PostgreSQL + pgvector, and serves it to AI agents through MCP, a CLI, and a REST API. Human-certified definitions drive answers, and when the retrieved candidates can't support a query it declines instead of guessing. Supports 10 databases. Apache-2.0 License, about 5.7k Stars.
Open Source Project #253: dsh-superagent — DeepSeek's Open-Source Proactive Coding Agent Prototype, Built Around Taking as Little Human Attention as Possible
dsh-superagent is a research prototype open-sourced by the DeepSeek team: a proactive coding agent for a single GitHub repository, packaged as a plugin bundle for DeepSeek Harness. It listens to GitHub (and optionally Feishu); when an allowed person asks, it opens a ticket, works in its own agent session and directory, and answers in the original conversation, for example with a pull request. Its design revolves around one sentence: take as little human attention as possible, and give as much value as possible. MIT licensed, 15k stars in three days. The README states plainly that it is not an official DeepSeek product.
One Open Source Project a Day (No. 252): Lightpanda — A Headless Browser Written From Scratch in Zig for AI Agents, 16x Less Memory and 9x Faster
Lightpanda is a headless browser written from scratch in Zig. It is not a Chromium fork or a WebKit patch, has no graphical rendering engine, and is built for AI agents and automation. In the project's own test on 933 real pages, 100 pages peaked at 123MB of memory (Headless Chrome - 2GB) and took 5 seconds (Chrome - 46 seconds). It speaks CDP so Puppeteer/Playwright can connect directly, and ships with an MCP server and a native agent. AGPL-3.0 License, about 40.7k Stars.
LLM-Driven Automated Testing Series (11): Why API Testing Is Actually the Most Conservative Battleground for LLMs
Keploy has 18.5k stars and its README says 'Expand API Coverage using AI' — but grep the entire open-source Go codebase for LLM/AI keywords and you'll find zero implementation code. Schemathesis, at 3.6k stars, uses no LLM at all — it finds real server 500 errors purely through Hypothesis-based property-based fuzzing. The projects that actually call an LLM API in their core logic turn out to be two sub-100-star projects: AutoRestTest and api-automation-agent. This article pins down why: structured API input/output is naturally suited to traditional fuzzing, so what can LLMs actually do here that traditional tools can't — and why does this domain stay cautious instead of going all-in the way UI automation has?
Open Source Project #251: Claude SEO — An SEO Audit Team Inside Claude Code, with 26 Sub-Skills and 19 Specialist Agents Checking Your Site in Parallel
Claude SEO is an open-source Claude Code plugin by AgriciDaniel. It splits technical SEO, E-E-A-T, schema, Core Web Vitals, GEO/AI search, and local and e-commerce SEO into 26 sub-skills and 19 specialist agents. One /seo audit command audits a whole site in parallel and returns a prioritized action plan. MIT licensed, about 23K stars.
Open Source Project #249: Hallmark — A Design Skill That Refuses to Look AI-Generated, Built on 48 Material Themes and a Browser Linter
Hallmark is a design skill from Together AI, open-sourced by Nutlope, for Claude Code, Codex, Cursor and OpenCode. It holds a hard floor (contrast, focus, reduced motion, no invented facts, no eyebrows), deals three themes from a catalog of 48 materials, builds the page around what the visitor came to do, lints the result in a real browser with sloplint, and hands the last word to a fresh-context reviewer. MIT licensed, nearly 38k stars.