Alvarmethod
One-to-one AI teaching skills (Alvar method) for Codex, Claude Code, Grok, Pi, and OpenCode
What is it?
What it is
Alvarmethod is a portable collection of agent skills that implement Eero Alvar’s AI learning loop (probe, plan, teach, lock‑in quiz) for personalized, agent‑driven learning.
Why it exists
It exists to eliminate many‑to‑many learning leaks by providing a single interface that draws from many sources, engineers trust through verification, keeps struggle in the material, and removes logistical overhead so learners focus on one reasoning step at a time.
Who should use it
Developers and power users of AI coding agents (Claude Code, Grok, Codex, Pi, OpenCode) who want to run Eero Alvar's AI learning loop for personalized, quiz-driven learning.
Who should avoid it
Users who prefer traditional static tutorials, those not using the skills CLI or the listed AI agents, and anyone looking for a simple one‑off learning resource.
How it works
A quick walkthrough in plain English
How Alvarmethod works
Step 1 of 3
You interact with it
Open Alvarmethod, send a request, or connect it to your stack.
Features
Advantages
- Truly cross-agent: one install works across Claude Code, Codex, Grok, Pi, OpenCode, and Cursor without per-agent rewrites
- Personalized learning: the probe → plan → teach loop tailors the DAG to the individual learner's edge rather than a fixed curriculum
- Lightweight distribution: a single Shell-based installer with no runtime dependencies beyond Node's npx
- Faithful to the original method while packaging it for modern coding-agent harnesses
- Quizzes use native UI pickers instead of in-chat A/B/C/D, which improves focus and reduces answer-hedging
- MIT license makes it easy to fork, embed, and redistribute
- Session files are human-readable Markdown, so learners can review progress in Obsidian or any editor
- Single-reasoning-step teaching reduces cognitive load and prevents textbook-dump behavior
- Adaptive prerequisite insertion automatically fills knowledge gaps when quizzes fail
- Clear separation between the open-source skill implementation and Eero Alvar's original video method
Disadvantages
- Documentation and skills are tightly coupled to one external video; if the link breaks or the method changes, context is lost
- npx-based install assumes Node.js and network access, which may not be available in restricted environments
- Quiz tooling depends on each agent exposing a native question picker — agents that lack one will degrade to in-chat MCQ
- Stars and adoption are modest (144 stars) with no listed topics, signaling a small community and limited peer review
- Most of the README is install instructions; there is little technical depth on how the mermaid DAG is actually generated
- Session state lives in a hidden `.alvar/` directory, which can be confusing for users who don't expect it inside their learning folder
- Contributing model requires editing skills in this repo and reinstalling — patching the installed copies under `~/.claude` or `~/.grok` is explicitly discouraged, slowing local iteration
- No automated tests are visible for the skills, so behavior changes may go undetected across agent harness updates
- Language is listed as Shell only, which understates the actual mix of Markdown skill definitions, SVG assets, and installer logic
Installation
native
## npx skills (preferred) Install via the open skills CLI, which discovers every `skills/*/SKILL.md` in the repo: ```bash # global, every detected agent, no prompts npx skills add vasanthsreeram/Alvarmethod -g --all # list what's in the pack npx skills add vasanthsreeram/Alvarmethod --list # one skill npx skills add vasanthsreeram/Alvarmethod --skill teach -g -y # one harness npx skills add vasanthsreeram/Alvarmethod -g -y -a claude-code -a grok -a pi -a opencode -a codex ``` Update later: ```bash npx skills update -g ``` ## npx this repo (no extra CLI) Runs `install.sh` directly from GitHub: ```bash npx github:vasanthsreeram/Alvarmethod npx github:vasanthsreeram/Alvarmethod --list npx github:vasanthsreeram/Alvarmethod --claude --grok ``` ## Clone ```bash git clone https://github.com/vasanthsreeram/Alvarmethod.git cd Alvarmethod ./install.sh # all known agent dirs ./install.sh --project ./install.sh --uninstall ``` Agent global skill paths installed by the installer: | Agent | Global path | |-------|-------------| | Codex | `~/.codex/skills` + `~/.agents/skills` | | Claude Code | `~/.claude/skills` | | Grok | `~/.grok/skills` | | Pi | `~/.pi/agent/skills` | | OpenCode | `~/.config/opencode/skills` |
FAQ
What is the Alvarmethod and what does it do?
Alvarmethod is a pack of portable Agent Skills that implement Eero Alvar's AI learning loop. It uses a probe → plan → teach → lock-in quiz workflow delivered as a slash command (e.g., /teach) inside AI coding agents. Rather than dumping a textbook, it probes the learner's edge, builds a mermaid DAG of prerequisites, teaches one reasoning step at a time, and locks each node in with a quiz.
How do I install the skills?
The preferred install is one line: `npx skills add vasanthsreeram/Alvarmethod -g --all`, which discovers every skills/*/SKILL.md in the repo and registers them globally for Claude Code, Codex, Grok, Pi, OpenCode, Cursor, and other skills-CLI agents. You can also run `npx github:vasanthsreeram/Alvarmethod` (no extra CLI needed, it executes install.sh), or clone the repo and run `./install.sh` directly. Update later with `npx skills update -g`.
How do I start a learning session?
Open a learning folder (not the repo itself) and run `/teach I want a solid introduction to <topic>`. The agent should respond with a native quiz picker (not A/B/C/D typed in chat), then a mermaid DAG, then teach one reasoning step, then a lock-in quiz. If the agent pastes choices in text, reply 'use the quiz tool'. If it dumps a textbook, reply 'one node only'.
What skills ship in the pack and what do they do?
Five skills ship: `teach` (the full probe → plan → one-step teaching loop), `probe` (builds the understanding map only), `learn-profile` (writes `.alvar/LEARNER.md` capturing the learner's profile), `learn-visual` (produces one SVG diagram you look at and refine), and `learn-verify` (fact-checks a claim before it is taught as fact).
Which quiz tool does each agent use, and why does it matter?
Quizzes must use the native picker, not text in chat. Grok and Codex call `ask_user_question`; Claude Code calls `AskUserQuestion`; OpenCode calls `question`; Pi uses `quiz`, `ask_user`, or `askUserQuestion`. Native pickers keep the graded MCQ flow and prevent the agent from hedging across many sources. If you see A/B/C/D pasted in the response, tell the agent to 'use the quiz tool'.
Where do session files live and can I use them in Obsidian?
Session artifacts are written to the learning folder under `.alvar/`: `LEARNER.md` for the learner profile, `maps/<topic>.md` for the understanding DAG, `sessions/<date>-<topic>.md` for transcripts, and `visuals/` for SVGs. Point Obsidian at the `.alvar/` folder if you want LaTeX preview of the rendered maps and notes.
Featured in Videos
YouTube tutorials and walkthroughs for Alvarmethod
Alternatives
Similar projects ranked by category, topics, and text overlap.
