nykooi1/vibe-wise

A Claude Code / Codex plugin that helps you learn how to build while AI writes the code.

What it solves

It addresses the problem of AI coding assistants writing too much code too quickly, which can leave developers—especially beginners or those in unfamiliar stacks—feeling disconnected from the design decisions and missing out on the learning process.

How it works

As a plugin for Claude Code, VibeWise shifts the workflow from immediate implementation to a guided design process. It introduces a series of "checkpoints" that force a dialogue between the user and the AI before any code is written:

  • Build Checkpoints: The AI asks the user to reason through the problem and approach.
  • Design Checkpoints: The user reviews the proposed architecture and confirms it or discusses alternatives.
  • Implementation Checkpoints: The user authorizes the specific code changes after the design is settled.

The plugin adapts its teaching style (Beginner, Intermediate, or Advanced) based on the user's experience level and maintains learning notes and project maps in a local .vibe-wise/ directory to preserve context across sessions.

Who it’s for

  • Aspiring engineers and junior developers who want to learn software design while building.
  • Experienced engineers working with an unfamiliar technology stack.
  • Anyone who wants to maintain strict ownership over architectural decisions while using AI to handle the boilerplate and implementation.

Highlights

  • Learning-First Workflow: Prioritizes planning and tradeoff analysis over instant code generation.
  • Adaptive Pedagogy: Adjusts the depth of explanations and the complexity of reasoning questions based on user level.
  • Customizable Friction: Users can adjust checkpoint frequency (Light, Normal, Frequent) to control how often the AI pauses for confirmation.
  • Local Context Management: Stores learning notes and system maps locally without requiring a separate backend or account.

Related

  • Project
    Farama-Foundation/PettingZoo

    A Python library for multi-agent reinforcement learning research that provides a standardized API and a diverse set of environments, similar to a multi-agent version of Gymnasium.

  • Project
    Mesh-LLM/mesh-llm

    A distributed LLM inference system that pools GPUs and memory across multiple machines to serve large models via an OpenAI-compatible API.

  • Project
    modelscope/evalscope

    A comprehensive LLM evaluation framework that combines capability benchmarking, inference performance stress testing, and interactive result visualization in one tool.

  • Project
    microsoft/hummingbird

    A library that compiles trained traditional ML models into tensor computations to accelerate inference using neural network frameworks like PyTorch.