FYI, updated
github.com/roderik/ai-... to auto install MCP and agents like for Claude Code
github.com/roderik/ai-... to auto install MCP and agents like for Claude Code
GitHub - roderik/ai-rules: Professional AI assistant configurations for Claude Code, Codex CLI, OpenCode and Gemini CLI with enterprise-grade defaults for Solidity and TypeScript development.
Professional AI assistant configurations for Claude Code, Codex CLI, OpenCode and Gemini CLI with enterprise-grade defaults for Solidity and TypeScript development. - roderik/ai-rules
github.com
August 20, 2025 at 7:36 PM
FYI, updated
github.com/roderik/ai-... to auto install MCP and agents like for Claude Code
github.com/roderik/ai-... to auto install MCP and agents like for Claude Code
Bonus, the sonic model (stealth) is currently free!
August 20, 2025 at 7:36 PM
Bonus, the sonic model (stealth) is currently free!
Bottom line: Structured comments are infrastructure for team knowledge.
They reduce cognitive load, preserve context, and turn code maintenance from detective work into informed decision-making.
Your team will thank you. Your 3 AM self definitely will.
They reduce cognitive load, preserve context, and turn code maintenance from detective work into informed decision-making.
Your team will thank you. Your 3 AM self definitely will.
August 18, 2025 at 8:00 AM
Bottom line: Structured comments are infrastructure for team knowledge.
They reduce cognitive load, preserve context, and turn code maintenance from detective work into informed decision-making.
Your team will thank you. Your 3 AM self definitely will.
They reduce cognitive load, preserve context, and turn code maintenance from detective work into informed decision-making.
Your team will thank you. Your 3 AM self definitely will.
Pro tip: Use comment tags for quick scanning:
- WHY: Core rationale
- TRADEOFF: Competing concerns
- SECURITY: Boundaries/implications
- PERF: Performance considerations
Makes code archaeology surgical instead of exploratory.
- WHY: Core rationale
- TRADEOFF: Competing concerns
- SECURITY: Boundaries/implications
- PERF: Performance considerations
Makes code archaeology surgical instead of exploratory.
August 18, 2025 at 8:00 AM
Pro tip: Use comment tags for quick scanning:
- WHY: Core rationale
- TRADEOFF: Competing concerns
- SECURITY: Boundaries/implications
- PERF: Performance considerations
Makes code archaeology surgical instead of exploratory.
- WHY: Core rationale
- TRADEOFF: Competing concerns
- SECURITY: Boundaries/implications
- PERF: Performance considerations
Makes code archaeology surgical instead of exploratory.
Key insight: Comments aren't documentation for the code. They're documentation for the DECISIONS that shaped the code.
"Why this algorithm?" "Why this trade-off?" "Why this error handling?"
Code reviews become conversations about choices, not syntax.
"Why this algorithm?" "Why this trade-off?" "Why this error handling?"
Code reviews become conversations about choices, not syntax.
August 18, 2025 at 8:00 AM
Key insight: Comments aren't documentation for the code. They're documentation for the DECISIONS that shaped the code.
"Why this algorithm?" "Why this trade-off?" "Why this error handling?"
Code reviews become conversations about choices, not syntax.
"Why this algorithm?" "Why this trade-off?" "Why this error handling?"
Code reviews become conversations about choices, not syntax.
The code-commenter agent I built scans TypeScript changes and adds structured comments automatically:
- TSDoc for exports
- Inline rationale for non-obvious blocks
- Focus on trade-offs, constraints, security boundaries
Consistency without the manual overhead.
- TSDoc for exports
- Inline rationale for non-obvious blocks
- Focus on trade-offs, constraints, security boundaries
Consistency without the manual overhead.
August 18, 2025 at 8:00 AM
The code-commenter agent I built scans TypeScript changes and adds structured comments automatically:
- TSDoc for exports
- Inline rationale for non-obvious blocks
- Focus on trade-offs, constraints, security boundaries
Consistency without the manual overhead.
- TSDoc for exports
- Inline rationale for non-obvious blocks
- Focus on trade-offs, constraints, security boundaries
Consistency without the manual overhead.
Decision context has a half-life measured in days.
August 18, 2025 at 8:00 AM
Decision context has a half-life measured in days.
Onboarding drops from weeks to days when context is embedded in code.
Your future self is also your teammate. That clever optimization you wrote 6 months ago? Without comments explaining the trade-offs, you'll spend an hour re-deriving why you chose that approach.
Your future self is also your teammate. That clever optimization you wrote 6 months ago? Without comments explaining the trade-offs, you'll spend an hour re-deriving why you chose that approach.
August 18, 2025 at 8:00 AM
Onboarding drops from weeks to days when context is embedded in code.
Your future self is also your teammate. That clever optimization you wrote 6 months ago? Without comments explaining the trade-offs, you'll spend an hour re-deriving why you chose that approach.
Your future self is also your teammate. That clever optimization you wrote 6 months ago? Without comments explaining the trade-offs, you'll spend an hour re-deriving why you chose that approach.
Context that saves your teammate 30 minutes of archaeology at 3 AM.
Team problem: New dev joins, spends 2 weeks learning "the way we do things" through trial and error.
Solution: Comments that preserve the WHY behind decisions.
Team problem: New dev joins, spends 2 weeks learning "the way we do things" through trial and error.
Solution: Comments that preserve the WHY behind decisions.
August 18, 2025 at 8:00 AM
Context that saves your teammate 30 minutes of archaeology at 3 AM.
Team problem: New dev joins, spends 2 weeks learning "the way we do things" through trial and error.
Solution: Comments that preserve the WHY behind decisions.
Team problem: New dev joins, spends 2 weeks learning "the way we do things" through trial and error.
Solution: Comments that preserve the WHY behind decisions.
Direct link:
github.com/roderik/ai-...
Here's why "why-first" comments are a game-changer for teams:
Bad comment: "// Increment counter"
Good comment: "// WHY: Debounce prevents API rate limit hits during search typing (5 QPS limit)"
The difference?
github.com/roderik/ai-...
Here's why "why-first" comments are a game-changer for teams:
Bad comment: "// Increment counter"
Good comment: "// WHY: Debounce prevents API rate limit hits during search typing (5 QPS limit)"
The difference?
ai-rules/.claude/agents/code-commenter.md at main · roderik/ai-rules
Professional Claude Code and Claude Desktop configuration with enterprise-grade defaults for Solidity and TypeScript development. - roderik/ai-rules
github.com
August 18, 2025 at 8:00 AM
Direct link:
github.com/roderik/ai-...
Here's why "why-first" comments are a game-changer for teams:
Bad comment: "// Increment counter"
Good comment: "// WHY: Debounce prevents API rate limit hits during search typing (5 QPS limit)"
The difference?
github.com/roderik/ai-...
Here's why "why-first" comments are a game-changer for teams:
Bad comment: "// Increment counter"
Good comment: "// WHY: Debounce prevents API rate limit hits during search typing (5 QPS limit)"
The difference?
Monorepo? No problem.
It detects workspace structure, creates per-package documentation, and sets up scoped Copilot instructions with applyTo globs.
Turborepo dependency graphs included.
Unlike fragmented /init commands, this creates a unified documentation system that every AI tool understands.
It detects workspace structure, creates per-package documentation, and sets up scoped Copilot instructions with applyTo globs.
Turborepo dependency graphs included.
Unlike fragmented /init commands, this creates a unified documentation system that every AI tool understands.
August 17, 2025 at 12:27 PM
Monorepo? No problem.
It detects workspace structure, creates per-package documentation, and sets up scoped Copilot instructions with applyTo globs.
Turborepo dependency graphs included.
Unlike fragmented /init commands, this creates a unified documentation system that every AI tool understands.
It detects workspace structure, creates per-package documentation, and sets up scoped Copilot instructions with applyTo globs.
Turborepo dependency graphs included.
Unlike fragmented /init commands, this creates a unified documentation system that every AI tool understands.
It uses context7 MCP and web search to fetch CURRENT best practices for your detected tech stack.
No more outdated advice. Always fresh, always relevant.
No more outdated advice. Always fresh, always relevant.
August 17, 2025 at 12:27 PM
It uses context7 MCP and web search to fetch CURRENT best practices for your detected tech stack.
No more outdated advice. Always fresh, always relevant.
No more outdated advice. Always fresh, always relevant.
What it creates:
• CLAUDE.md - Project overview, stack, commands, best practices
• GEMINI.md, AGENTS.md - Symlinked to Claude docs
• copilot-instructions.md - Unified guidance
• Per-package docs for monorepos
The magic is in the research phase.
• CLAUDE.md - Project overview, stack, commands, best practices
• GEMINI.md, AGENTS.md - Symlinked to Claude docs
• copilot-instructions.md - Unified guidance
• Per-package docs for monorepos
The magic is in the research phase.
Adding custom instructions for GitHub Copilot - GitHub Docs
copilot-instructions.md
August 17, 2025 at 12:27 PM
What it creates:
• CLAUDE.md - Project overview, stack, commands, best practices
• GEMINI.md, AGENTS.md - Symlinked to Claude docs
• copilot-instructions.md - Unified guidance
• Per-package docs for monorepos
The magic is in the research phase.
• CLAUDE.md - Project overview, stack, commands, best practices
• GEMINI.md, AGENTS.md - Symlinked to Claude docs
• copilot-instructions.md - Unified guidance
• Per-package docs for monorepos
The magic is in the research phase.
It analyzes your repository and generates comprehensive documentation that works across ALL your AI tools - Claude, Gemini, Copilot, and more.
One source of truth. No more context confusion.
One source of truth. No more context confusion.
August 17, 2025 at 12:27 PM
It analyzes your repository and generates comprehensive documentation that works across ALL your AI tools - Claude, Gemini, Copilot, and more.
One source of truth. No more context confusion.
One source of truth. No more context confusion.
Installation is stupid simple:
```bash
curl -sL raw.githubusercontent.com/roderik/she... | bash
```
That's it. Installs Homebrew, all tools, configures everything.
No manual setup. No broken configs. Just works.
```bash
curl -sL raw.githubusercontent.com/roderik/she... | bash
```
That's it. Installs Homebrew, all tools, configures everything.
No manual setup. No broken configs. Just works.
August 16, 2025 at 8:58 AM
Installation is stupid simple:
```bash
curl -sL raw.githubusercontent.com/roderik/she... | bash
```
That's it. Installs Homebrew, all tools, configures everything.
No manual setup. No broken configs. Just works.
```bash
curl -sL raw.githubusercontent.com/roderik/she... | bash
```
That's it. Installs Homebrew, all tools, configures everything.
No manual setup. No broken configs. Just works.
Smart functions included:
• Claude wrapper with auto-detection
• Docker management shortcuts
• Git branch cleanup utilities
• Directory jumping with zoxide
• History search with atuin
All modular. All documented. All tested.
• Claude wrapper with auto-detection
• Docker management shortcuts
• Git branch cleanup utilities
• Directory jumping with zoxide
• History search with atuin
All modular. All documented. All tested.
August 16, 2025 at 8:58 AM
Smart functions included:
• Claude wrapper with auto-detection
• Docker management shortcuts
• Git branch cleanup utilities
• Directory jumping with zoxide
• History search with atuin
All modular. All documented. All tested.
• Claude wrapper with auto-detection
• Docker management shortcuts
• Git branch cleanup utilities
• Directory jumping with zoxide
• History search with atuin
All modular. All documented. All tested.
The magic is in the details:
`ls` → `eza -lh --group-directories-first`
`cat` → `bat` (syntax highlighting)
`grep` → `ripgrep` (10x faster)
`find` → `fd` (actually usable)
Same experience across all three shells.
`ls` → `eza -lh --group-directories-first`
`cat` → `bat` (syntax highlighting)
`grep` → `ripgrep` (10x faster)
`find` → `fd` (actually usable)
Same experience across all three shells.
August 16, 2025 at 8:58 AM
The magic is in the details:
`ls` → `eza -lh --group-directories-first`
`cat` → `bat` (syntax highlighting)
`grep` → `ripgrep` (10x faster)
`find` → `fd` (actually usable)
Same experience across all three shells.
`ls` → `eza -lh --group-directories-first`
`cat` → `bat` (syntax highlighting)
`grep` → `ripgrep` (10x faster)
`find` → `fd` (actually usable)
Same experience across all three shells.
• Triple shell support (Fish/Zsh/Bash)
• 60+ git aliases that actually make sense
• Neovim with LazyVim config
• 60+ git aliases that actually make sense
• Neovim with LazyVim config
August 16, 2025 at 8:58 AM
• Triple shell support (Fish/Zsh/Bash)
• 60+ git aliases that actually make sense
• Neovim with LazyVim config
• 60+ git aliases that actually make sense
• Neovim with LazyVim config
The problem: Default terminals are garbage. Default shells are worse.
You spend 8+ hours a day in there but it looks like 1995 and acts like it too.
What you get:
• Starship prompt (context-aware, blazing fast)
• Modern tools: bat, eza, ripgrep, fd, fzf
You spend 8+ hours a day in there but it looks like 1995 and acts like it too.
What you get:
• Starship prompt (context-aware, blazing fast)
• Modern tools: bat, eza, ripgrep, fd, fzf
August 16, 2025 at 8:58 AM
The problem: Default terminals are garbage. Default shells are worse.
You spend 8+ hours a day in there but it looks like 1995 and acts like it too.
What you get:
• Starship prompt (context-aware, blazing fast)
• Modern tools: bat, eza, ripgrep, fd, fzf
You spend 8+ hours a day in there but it looks like 1995 and acts like it too.
What you get:
• Starship prompt (context-aware, blazing fast)
• Modern tools: bat, eza, ripgrep, fd, fzf