Daily cognitive coding challenges inside Neovim. Fight AI-induced brain atrophy — solve data structure and algorithm problems without Copilot, without autocomplete, just you and the code.

The more you delegate to AI, the more your problem-solving muscles atrophy. brain.nvim gives you one challenge per day — a split view with a TypeScript stub on the left, a chat panel on the right, and auto-running tests on save. No AI assistance unless you explicitly ask for a hint (which costs you points).

• 8 built-in challenges — LRU Cache, Trie, Debounce, Deep Clone, Event Emitter, Promise.all, Binary Search, Flatten Array
• Deterministic rotation — one challenge per day, cycles through all before repeating
• Split-view UI — code on the left, chat/feedback on the right
• Auto-test on save — vitest runs every time you save your challenge file
• Scoring — ✅ 3pts (clean solve), 🟡 2pts (with hints), 🔴 1pt (partial), ⏭️ 0pts (skip)
• Training log — Markdown table tracking your scores over time
• Copilot/completion auto-disabled in challenge buffers
• Optional AI hints — via any OpenAI-compatible API (OpenClaw, LM Studio, etc.)
• Chat panel — discuss the problem with an AI tutor (Socratic style, no freebies)

{
  'execute008/brain.nvim',
  cmd = { 'BrainStart', 'BrainTest', 'BrainHint', 'BrainSubmit', 'BrainSkip' },
  opts = {
    -- Where challenge files are created (one folder per day)
    challenges_dir = vim.fn.expand('~/.local/share/brain.nvim/challenges'),

    -- Markdown log of your scores
    log_file = vim.fn.expand('~/.local/share/brain.nvim/training.md'),

    -- Test runner (must support --reporter=json)
    test_cmd = { 'npx', 'vitest', 'run', '--reporter=json' },

    -- Disable Copilot/completion in challenge buffers
    disable_copilot = true,

    -- AI hints: "openclaw" or "none"
    hint_provider = 'none',

    -- OpenAI-compatible API for hints/chat (any local or remote endpoint)
    openclaw = {
      port = 18789,
      token = '',
    },

    -- Callback fired on submit (for custom integrations)
    -- on_submit = function(result) ... end,
  },
}

The challenges use TypeScript + vitest. After installing, run once:

cd ~/.local/share/brain.nvim/challenges
npm install

(brain.nvim creates a package.json automatically if one doesn't exist.)

Results are logged to your log_file in a Markdown table:

| Date | Challenge | Score | Tests | Hints |
|------|-----------|-------|-------|-------|
| 2026-02-18 | Trie (Prefix Tree) | ✅ 3/3 | 11/11 | 0 |
| 2026-02-19 | LRU Cache | 🟡 2/3 | 9/9 | 1 |

brain.nvim can connect to any OpenAI-compatible chat endpoint for:

• Hints (:BrainHint) — brief nudges without spoiling the answer
• Chat — Socratic coaching in the side panel
• Skip explanations (:BrainSkip) — full solution walkthrough

Works with OpenClaw, LM Studio, Ollama, or any /v1/chat/completions endpoint.

opts = {
  hint_provider = 'openclaw',
  openclaw = {
    port = 11434,  -- or whatever your endpoint uses
    token = 'your-token',
  },
}

Want to send results to Telegram, Discord, or your own tracking system?

opts = {
  on_submit = function(result)
    -- result = { date, challenge, score, emoji, hints, passed, total }
    vim.notify(result.emoji .. ' ' .. result.challenge .. ': ' .. result.score .. '/3')
  end,
}

Edit lua/brain/challenges_db.lua or submit a PR! Each challenge needs:

• name — display name
• difficulty — "easy", "medium", or "hard"
• stub — TypeScript template with // YOUR CODE HERE markers
• tests — vitest test suite

The goal isn't to grind LeetCode. It's to keep your brain sharp while AI handles the routine stuff. One problem a day, no shortcuts, real thinking.

MIT