2026-10-21
Introduction
Love using AI to write code and happily pushing it, only to realize you don't actually understand what it does? Sphinx, an open-source tool trending on Hacker News this week, intercepts you right before your Git push, forcing you to take a code comprehension quiz and making you be a responsible engineer! I'm Muyan. In this week's "Open Source Weekly," besides taking this foolproof tool for a spin, I'll also introduce a cutting-edge tool that simulates LLM speeds on various GPUs right in your browser. Want to know how fast your Mac can run Llama 3? I'll tell you in a bit!
This Week's Open Source Picks
1. Sphinx: An AI Code Comprehension Quiz Tool Before Git Push
- What it is: It solves a major pain point for modern engineers: over-relying on AI to write code and blindly pushing it. Once you finish writing code with AI assistance, it inserts a quiz right before your Git push to ensure you actually understand the AI-generated code, rather than blindly submitting a PR.
- Buzz this week: It gained significant traction on Hacker News with 252 points, sparking 89 comments in a heated debate about "how to maintain code quality in the AI era."
- Key Features:
- Git Hook Interception and Automated Quizzes: It automatically intercepts your workflow during the Git push phase, reads your commit diff, and uses an LLM to generate a comprehension quiz.
- Support for Multiple AI Assistants: It perfectly supports code generated by mainstream AI coding assistants like Claude Code, Cursor, and Copilot.
- Custom LLM Endpoints: It supports OpenRouter or any OpenAI API-compatible service, eliminating the need for extra paid subscriptions to specific SaaS platforms.
- YOLO Escape Hatch: If you are in a rush to deploy, it supports using
yoloto quickly bypass the quiz, but every bypass is recorded in the logs for future team audits, leaving a paper trail if you try to slack off.
- Quick Start:
Install it globally and bind it to your Git hooks in your project directory:
npm install -g sphinx-understand sphinx install - How it compares: Unlike traditional static analysis tools or automated tests, Sphinx doesn't just check code syntax or logic. Instead, it directly tests the "human developer's" understanding of the code.
- Who it's for: Perfect for engineers who want to break the habit of "blindly trusting AI," or team leads who want to ensure team members take responsibility for their own code.
- Link: Sphinx Official Website
2. Llmspeed: An LLM Execution Speed Simulator in the Browser
- What it is: Designed for engineers who want to run open-source LLMs locally but struggle with hardware choices. It simulates the actual generation speed of specific models on different GPUs or Mac chips directly in the browser, solving the pain point of buying the wrong hardware and wasting money.
- Buzz this week: It garnered 329 points on Hacker News, attracting 86 comments.
- Key Features:
- Hardware Speed Simulation: It simulates the tokens per second (tokens/sec) when running specific open-source models on various mainstream hardware, such as the NVIDIA RTX series and Apple Silicon Macs.
- Perceived Speed Visualization: It simulates Time to First Token (TTFT), allowing users to experience the actual perceived speed of the "typewriter" output.
- Community-Backed Data: The data is based on community-verified benchmarks, and users can submit new hardware data via Pull Requests.
- Quick Start:
You can use the web version directly by opening
llmspeed.dev. If you want to run it locally, clone the repository and run:npm install && npm run dev - How it compares: Unlike benchmark tables that only provide cold performance numbers (such as LLM-Perf), Llmspeed provides an intuitive simulation of perceived speed by rendering the text stream in real time right in the browser.
- Who it's for: Ideal for engineers facing decision paralysis over which graphics card or Mac to buy for running local LLMs.
- Link: Llmspeed Official Website
3. Esp32sim: A Pure Rust Browser-Based ESP32 Instruction-Level Simulator
- What it is: It allows embedded and IoT developers to simulate ESP32 chips directly in the browser, solving the pain point of not having a physical development board on hand or wanting to quickly validate firmware logic.
- Buzz this week: It received 206 points on Hacker News (with 46 comments) and currently has 545 stars on GitHub under the MIT license.
- Key Features:
- Dual-Architecture Instruction-Level Simulation: It supports both Xtensa (ESP32, S3) and RISC-V (C3, C6) architectures.
- Full ROM Booting: It can boot the actual mask ROM, bootloader, and unmodified application images directly in the browser via WebAssembly.
- Peripheral and UI Simulation: It supports simulating Touch-LCD screens, smoothly rendering LVGL UIs with mouse interaction support.
- Quick Start:
You can experience the web version directly by opening
mikroverk.github.io/esp32sim/. For local execution, you can run it via Rust:cargo run - How it compares: Compared to online simulators like Wokwi, Esp32sim is an open-source, instruction-level simulator written entirely in Rust. It simulates the actual ROM boot process at a lower level and operates completely independent of cloud servers.
- Who it's for: Perfect for IoT engineers who want to avoid constantly plugging and unplugging USB cables and want to quickly debug ESP32 firmware and UIs in the browser.
- Link: Esp32sim GitHub
Closing Thoughts
From foolproof Git quizzes and LLM speed simulations to in-browser chip simulators, this week's tools significantly reduce our development and decision-making costs. I highly recommend everyone give Sphinx a try. After all, forcing yourself to understand AI-written code is the ultimate way to secure your job. See you in next week's "Open Source Weekly"!
Comments