diff options
Diffstat (limited to 'docs/NOTES.org')
| -rw-r--r-- | docs/NOTES.org | 583 |
1 files changed, 0 insertions, 583 deletions
diff --git a/docs/NOTES.org b/docs/NOTES.org deleted file mode 100644 index a08a25e8..00000000 --- a/docs/NOTES.org +++ /dev/null @@ -1,583 +0,0 @@ -#+TITLE: 🚨 ACTIVE PROJECT - READ THIS FIRST 🚨 -#+AUTHOR: Claude Code Session Notes -#+DATE: 2025-10-30 - -* 🗣️ IMPORTANT TERMINOLOGY - -** "I want to do an X session with you" - -When Craig says "I want to do an X session with you", this means: -- **CREATE a session definition** for doing X (meta-work) -- **NOT** "let's DO X right now" (the actual work) - -This triggers the create-session workflow from docs/sessions/create-session.org - -*Examples:* -- "I want to do an emacs inbox zero session" → Create docs/sessions/inbox-zero.org -- "I want to do a refactor session" → Create docs/sessions/refactor.org -- "I want to do a code review session" → Create docs/sessions/code-review.org - -* 🔔 DESKTOP NOTIFICATIONS WORKFLOW - -**IMPORTANT: How Claude notifies you when blocked** - -When Claude needs your input on blocking questions, Claude will send a desktop notification via `notify-send`: - -#+BEGIN_SRC bash -notify-send "Claude Code" "Question: [Your input needed]" --urgency=normal -#+END_SRC - -**When notifications ARE sent:** -- ✅ When explicitly needing your decision/input (blocking questions) -- ✅ When multiple valid approaches exist and choice affects implementation -- ✅ When encountering errors that require user guidance -- ✅ **ONLY** when Claude cannot proceed without user input - -**When notifications are NOT sent:** -- ❌ After completing tasks (informational updates) -- ❌ During normal progress updates -- ❌ When milestones are reached -- ❌ For status messages or completion notices -- ❌ ANY informational alerts - -**Setup:** -- Requires `dunst` or similar notification daemon -- Works with `notify-send` command -- Always uses `--urgency=normal` (not critical) - -**Purpose:** -This allows you to context-switch to other work while Claude runs long tasks, and get notified ONLY when your input is truly needed to continue. You check back when convenient for status updates. - -* 🧩 EMACS LISP DEVELOPMENT BEST PRACTICES - -**Critical Lessons: Preventing Parenthesis Errors** - -Both humans and AI struggle with balanced parentheses in deeply nested Emacs Lisp code. Here are proven strategies to prevent this: - -** For AI Code Generation - -*** 1. Write Small, Focused Functions -- Keep functions under 15 lines when possible -- Each function should do ONE thing -- Easier to verify parentheses at a glance -- Easier to test in isolation - -#+BEGIN_EXAMPLE -Bad (deeply nested, hard to verify): -(defun process-data (data) - (when (valid-p data) - (let ((result (transform data))) - (when result - (let ((final (format result))) - (when final - (save final))))))) - -Good (broken into helpers): -(defun process-data (data) - (when (valid-p data) - (save-result (format-result (transform-data data))))) - -(defun transform-data (data) ...) -(defun format-result (result) ...) -(defun save-result (final) ...) -#+END_EXAMPLE - -*** 2. Test Immediately After Each Write -- Write function → check-parens → test load → repeat -- Don't batch multiple functions before testing -- Catch errors early when context is fresh - -#+BEGIN_SRC bash -# After writing each function: -emacs --batch file.el --eval '(check-parens)' && echo "✓" -#+END_SRC - -*** 3. Prefer Write Over Multiple Edits -- For complex new code: use Write tool (complete file) -- Only use Edit for small, simple changes -- Incremental edits can introduce subtle paren mismatches -- Complete file Write is easier to verify - -*** 4. Validate Before Committing -- Use pre-commit hooks to validate all .el files -- Prevents committing broken code -- Example in chime.el repository: .git/hooks/pre-commit - -** For Human Developers - -*** 1. Use Structural Editing Modes -These PREVENT unbalanced parens: -- **paredit** - Classic, strict structural editing -- **smartparens** - More flexible, works with multiple languages -- **lispy** - Modal editing for lisps -- **electric-pair-mode** - Built-in, auto-closes parens - -*** 2. Enable Real-time Validation -- **flycheck** + **flycheck-package** - Shows errors as you type -- **flymake** - Built-in alternative -- **rainbow-delimiters-mode** - Colors matching parens - -*** 3. Quick Validation Commands -#+BEGIN_SRC elisp -M-x check-parens ; Check current buffer -M-x byte-compile-file ; More comprehensive checking -#+END_SRC - -** Lessons from chime-org-contacts.el Development - -*** Original Problem -Complex nested function with 6 levels of nesting: -- Hard to count parentheses manually -- AI kept adding/removing wrong number of closing parens -- Functions weren't defined after file load -- Wasted significant debugging time - -*** Solution That Worked -Refactored into 3 helper functions: -1. =chime-org-contacts--parse-birthday= (10 lines) -2. =chime-org-contacts--format-timestamp= (4 lines) -3. =chime-org-contacts--insert-timestamp-after-drawer= (12 lines) - -Main function became simple composition (10 lines). - -**Result:** -- All functions defined correctly -- Easy to verify parens by eye -- Each function testable independently -- 24 tests written covering all cases - -*** Key Insight -Breaking complex code into small helpers: -- ✅ Easier to verify correctness -- ✅ Easier to test -- ✅ Easier to maintain -- ✅ Self-documenting through function names -- ✅ AI and humans both succeed - -Deeply nested code: -- ❌ Hard to verify -- ❌ Hard to test -- ❌ AI frequently makes paren errors -- ❌ Humans make mistakes too - -** Tools and Workflow Summary - -| Stage | Tool | Purpose | -|-------+------+---------| -| Writing | paredit/smartparens | Prevent errors | -| Editing | rainbow-delimiters | Visual verification | -| Testing | check-parens | Quick syntax check | -| CI/CD | pre-commit hooks | Prevent bad commits | -| Review | byte-compile-file | Comprehensive check | - -** Makefile - Comprehensive Testing & Validation Tool - -A comprehensive Makefile is available in the repository root with targets for testing, validation, and utilities. - -#+BEGIN_SRC bash -make # Show all available targets -make test # Run all tests (unit + integration) -make test-file FILE=... # Run specific test file -make test-name TEST=... # Run tests matching pattern -make validate-parens # Check for unbalanced parentheses -make validate-modules # Load all modules to verify compilation -make compile # Byte-compile all modules -make lint # Run checkdoc, package-lint, elisp-lint -make profile # Profile Emacs startup -make clean # Remove test artifacts and compiled files -make reset # Reset to first launch (destructive!) -#+END_SRC - -Created: 2025-11-03 -Adapted from chime.el Makefile with config-specific enhancements - -* 📋 AVAILABLE SESSION TYPES - -** create-session -File: [[file:sessions/create-session.org][docs/sessions/create-session.org]] - -Meta-workflow for creating new session types. Use this when identifying repetitive workflows that would benefit from documentation. - -Workflow: -1. Q&A discovery (4 core questions) -2. Assess completeness -3. Name the session -4. Document it -5. Update NOTES.org -6. Validate by execution - -Created: 2025-11-01 (pre-existing) - -** emacs-inbox-zero -File: [[file:sessions/emacs-inbox-zero.org][docs/sessions/emacs-inbox-zero.org]] - -Weekly workflow for processing the "Emacs Config Inbox" heading in =todo.org= to zero by filtering through V2MOM framework. - -Workflow: -1. Sort by priority (A → B → C → none → D) -2. Claude rereads V2MOM -3. Process each item through 3 questions: - - Does this need to be done? → DELETE if no - - Related to V2MOM? → Move to someday-maybe if no - - Which method? → Move to appropriate method -4. Done when inbox heading is empty - -Target: 10 minutes active work time -Cadence: Every Sunday, no longer than 7 days between sessions -Maintains metrics: Active todos < 20, weekly triage consistency - -Created: 2025-11-01 - -* CURRENT PROJECT STATUS - -** 🎯 What We're Doing -Working through a systematic approach to clean up and prioritize Craig's Emacs config work: - -1. ✅ *COMPLETE V2MOM* (Vision, Values, Methods, Obstacles, Metrics) - IN PROGRESS -2. ⏳ *TRIAGE todo.org* - Use V2MOM to ruthlessly cancel ~60% of tasks -3. ⏳ *EXECUTE TIER 1* - Ship quick wins (network check removal, Corfu, bug fixes) -4. ⏳ *BUILD OBSERVABILITY* - Create profiling infrastructure (TIER 2) -5. ⏳ *SYSTEMATIC EXECUTION* - Work through prioritized tasks one by one - -** 📍 Where We Are Right Now -*Session Started:* 2025-10-30 -*Current Step:* ✅ V2MOM COMPLETE - Ready for execution -*Time Committed:* ~2 sessions, V2MOM finished 2025-10-31 -*Status:* V2MOM complete, ready to begin Method 1 execution - -** 📄 Key Documents - -*** Primary Working Documents -- *V2MOM:* [[file:EMACS-CONFIG-V2MOM.org][EMACS-CONFIG-V2MOM.org]] - Strategic framework for Emacs config (✅ COMPLETE) - - Vision, Values, Methods, Obstacles, Metrics - - Used for decision-making and weekly triage - - Read this first to understand strategic direction -- *Issues Analysis:* [[file:../issues.org][../issues.org]] - Claude's detailed analysis with TIER system and implementations -- *Current Inbox:* [[file:../inbox.org][../inbox.org]] - V2MOM-aligned tasks (~23 items after ruthless triage) - -*** Reference Documents -- *Config Root:* [[file:../init.el][../init.el]] -- *Modules:* [[file:../modules/][../modules/]] -- *Tests:* [[file:../tests/][../tests/]] - -** 🔑 Key Insights About Craig's Work Patterns - -*** Strengths -- Thoughtful and strategic thinker -- Good research skills (thorough specs, complete code examples) -- Does ship things (dashboard, dirvish, network check fixes) -- Recognizes need for V2MOM framework -- Uses config daily for real work - -*** Patterns to Address -1. *Research > Execution* - Has complete code for Corfu, difftastic, transcription workflow... still TODO -2. *Priority Inflation* - Too many [#A]/[#B] items, unclear what's actually urgent -3. *Incomplete Strategy* - V2MOM structure exists but sections are empty -4. *Hard to Say No* - [#C]/[#D] items should be CANCELLED but remain in list -5. *Side Projects Compete* - Dupre theme work competes with core config maintenance - -*** What Craig Told Us About Himself -> "I am building tools both because they solve problems, but also because I enjoy building." - -This is healthy! But need balance: -- Fix rough edges FIRST (daily pain points) -- Build fun stuff SECOND (after maintenance) -- Cancel distractions ALWAYS (Signal client, minimap, etc.) - -** 🎯 Agreed Goals for This Project - -*** Immediate (Next 2-3 Sessions) -1. ✅ Complete V2MOM (IN PROGRESS) -2. ⏳ Triage todo.org using V2MOM as filter -3. ⏳ Execute quick wins: network check, Corfu migration, bug fixes -4. ⏳ Build debug-profiling.el infrastructure - -*** Short Term (Next Month) -5. Profile and optimize org-agenda performance -6. Ship reveal.js presentation workflow -7. Establish weekly triage ritual - -*** Long Term (Ongoing) -8. Ship more than research -9. Maintain < 20 active todos -10. Measure metrics from V2MOM - -** 📋 TIER System from issues.org - -*** TIER 1: Do These First (High Impact, Low Effort) - 1 weekend -- Remove network check (15 min) -- Fix missing functions (30 min) -- Corfu migration (2 hours) -- Mood-line switch (30 min) -- Bug fixes (1 hour) - -*** TIER 2: Build Observability (HIGHEST VALUE) - 1 week -- Create debug-profiling.el module (3-4 hours) -- Profile org-agenda-rebuild (1 hour) -- Add instrumentation and caching (2 hours) -- Test org-agenda filtering functions (2-3 hours) - -*** TIER 3: Quick Wins (After Profiling) - 1-2 hours each -- Reveal.js presentation workflow (2 hours) -- Difftastic integration (30 min) -- Local package development workflow (1 hour) - -*** TIER 4: Maybe/Someday (Probably Never) -- Code-maat reimplementation (HOLD) -- LaTeX config (HOLD until concrete need) -- Elfeed dashboard (HOLD - unclear if actually used) -- DWIM shell integration (HOLD - current solution works) -- Jumper package (HOLD - already maintaining chime + org-msg) - -** 🚫 Items That Should Be CANCELLED - -From todo.org, these don't serve the vision: -- [#D] Signal Client - Not in vision -- [#D] Awesome-tray / mode-icons - Already have modeline -- [#C] Minimap - Interesting, not important -- [#C] Install Magit TODOs - Already works fine -- [#C] Git Timemachine litters buffers - Minor annoyance -- Many Dupre theme TODOs - Side project competing with maintenance - -## 💡 Key Recommendations for Craig - -### Week 1: Strategy + Quick Wins -1. Complete V2MOM (2-3 hours) -2. Triage todo.org using V2MOM (1-2 hours) -3. Execute items you already have code for (2-3 hours) - -### Week 2: Observability Infrastructure -4. Build debug-profiling.el (3-4 hours) -5. Profile org-agenda (1 hour) - -### Week 3: Fix Performance + Ship Presentation -6. Fix org-agenda based on profiling (2-3 hours) -7. Ship reveal.js workflow (2 hours) - -### Ongoing: Maintenance Discipline -- Weekly triage ritual (30 min every Sunday) -- Measure metrics (startup time, agenda time, todo count) -- Ship > Research - -** 🔄 Next Session Pickup Points - -When starting next session, Claude should: - -1. **Read this document first** to understand context -2. **Check V2MOM status** - If incomplete, continue there -3. **Reference issues.org** for detailed technical recommendations -4. **Reference todo.org** for items to triage -5. **Ask Craig:** "Where did we leave off? V2MOM? Triage? Execution?" - -** 📞 Questions to Ask Craig Next Session - -*IMMEDIATE (when resuming):* -- "Ready to continue V2MOM? We left off at Methods section." -- "How much time do you have?" - -*FOR METHODS SECTION:* -Show Craig the draft list and ask: -- "Which methods do you already do consistently?" -- "Which do you want to do but don't yet?" -- "Am I missing any important methods?" - -*AFTER V2MOM COMPLETE:* -- "Ready to triage todo.org using the V2MOM?" -- "Should we execute quick wins or continue systematic triage?" - -** 🎯 Success Metrics for This Project - -We'll know this is working when: -- ✅ V2MOM is complete and provides clear strategic direction -- ✅ todo.org shrinks from ~50 to < 20 active items -- ✅ Craig ships 3-5 items per week (small but consistent) -- ✅ Craig has profiling infrastructure to measure performance -- ✅ Org agenda rebuild time is measured and improving -- ✅ Weekly triage becomes habit - -** 💬 Craig's Words to Remember - -> "I think you should adjust issues.org with all your recommendations. They are exciting, eye-opening, and just feel right. Add even your guidance on latex. spot on. thanks for your honesty. I did ask for it and am genuinely grateful for your responses. I'll take action on them." - -> "What I need help with is integrating this in with my existing todo.org file... Some of the tasks I've listed should probably just be deleted or better yet, marked CANCELLED." - -> "I have about an hour to devote. You could lead me through it, I could do some questions/answer rounds with you to clarify my thinking." - -Craig is ready to execute. He asked for honesty and took it well. He recognizes the patterns and wants systematic help. - -** 🛠️ Technical Context - -*** Current Pain Points -1. Org agenda is slow (performance bottleneck) -2. Network check adds 1+ seconds to startup (technical debt) -3. Missing functions cause errors (cj/log-silently, cj/goto-git-gutter-diff-hunks) -4. Mail attachments workflow is awkward -5. No profiling infrastructure to measure performance - -*** Items Craig Already Has Code For -These can be executed immediately - just paste and test: -- Transcription workflow (complete bash + elisp in todo.org:2-99) -- Difftastic integration (complete config in todo.org:1211-1223) -- Corfu migration (complete config in todo.org:1611-1639) - -*** Architecture -- Modular structure: modules/*.el -- Good test coverage for utilities -- Modern packages: Vertico/Consult/Embark stack -- Local package development: chime.el, org-msg - -** 📚 Related Reading - -If Craig or Claude need more context: -- [[file:../issues.org::*Second Opinion: Ruthless Prioritization & Reality Checks][Second Opinion section in issues.org]] - Full analysis and recommendations -- [[file:../issues.org::*TIER 1: Do These First][TIER 1-4 breakdown]] - Prioritized task system -- [[file:../quality-engineer.org][quality-engineer.org]] - Testing philosophy (if exists) - -** 🚀 Current Session Notes - -*** 2025-10-31 Session 2 - V2MOM Complete! -*Time:* ~1.5 hours -*Status:* ✅ COMPLETE - V2MOM finalized and ready for use - -*What We Completed:* -1. ✅ Finalized all 6 Methods with aspirational bodies and concrete actions: - - Method 1: Make Using Emacs Frictionless (performance & functionality fixes) - - Method 2: Stop Problems Before They Appear (proactive package maintenance) - - Method 3: Make *Fixing* Emacs Frictionless (observability/tooling) - - Method 4: Contribute to the Emacs Ecosystem (package maintenance tooling) - - Method 5: Be Kind To Your Future Self (new features) - - Method 6: Develop Disciplined Engineering Practices (meta-method with measurable outcomes) - -2. ✅ Completed Obstacles section (6 honest, personal obstacles with real stakes) - - Building vs fixing tension - - Getting irritated at mistakes - - Hard to say "no" - - Perfectionism - - Limited time sessions - - New habits are hard to sustain - -3. ✅ Completed Metrics section (Performance, Discipline, Quality metrics) - - Startup time: < 3s (currently 6.2s) - - Org-agenda: < 5s (currently 30+s) - - Active todos: < 20 (currently ~50+) - - Weekly triage consistency - - Research:shipped ratio > 1:1 - - Config uptime: never broken > 2 days - - Test coverage: > 70% with justification for uncovered code - -4. ✅ Implemented cj/diff-buffer-with-file - - Added to modules/custom-buffer-file.el - - Bound to C-; b D - - Unified diff format with proper error handling - - TODO comment for future difftastic integration - -5. ✅ Added missing items to Methods based on Craig's research: - - Fixed org-noter (Method 1) - - Added Buttercup (Method 3) - - Added package maintenance tools (Method 4: package-lint, melpazoid, elisp-check, undercover) - - Added wttrin to maintained packages list - -*Key Insights:* -- Craig wants to DELETE research files: "There will always be cool ideas out there to implement and they will always be a web search away" -- Ruthless prioritization is already happening -- Method ordering: fix → stabilize → build infrastructure → contribute → enhance → sustain -- Adjusted startup target from 2s to 3s (more achievable, less perfectionism trap) - -*Key Files Modified This Session:* -- [[file:emacs-config-v2mom.org][emacs-config-v2mom.org]] - Now 100% complete with all sections filled -- [[file:../modules/custom-buffer-file.el][modules/custom-buffer-file.el]] - Added cj/diff-buffer-with-file function -- [[file:SESSION-HANDOFF-ACTIVE-PROJECT.org][SESSION-HANDOFF-ACTIVE-PROJECT.org]] - This file - -*Next Session Starts With:* -1. Continue Method 1 execution - 2 quick wins ready! -2. Fix cj/goto-git-gutter-diff-hunks (15 min) -3. Fix chime throw/catch bug (your package) -4. Fix go-ts-mode-map keybinding error - -*** 2025-10-31 Session 3 - RUTHLESS EXECUTION! 🚀 -*Time:* ~2 hours -*Status:* Method 1 in progress - shipped 2 wins, discovered 3 bugs - -*What We Completed:* -1. ✅ **RUTHLESS PRIORITIZATION EXECUTED!** - - Moved todo.org → docs/someday-maybe.org (~50 items archived) - - Created fresh inbox.org with ONLY V2MOM-aligned tasks (23 items) - - Already under < 20 active items goal! - -2. ✅ **SHIPPED: Network check removal** (Method 1) - - Deleted `cj/internet-up-p` blocking ping function - - Removed network cache variables - - Simplified to use package priorities instead - - .localrepo (priority 200) ensures offline reproducibility - - **RESULT: 6.19s → 4.16s startup time (2.03 seconds faster!)** - -3. ✅ **SHIPPED: cj/diff-buffer-with-file** (Method 1) - - Implemented in modules/custom-buffer-file.el - - Bound to C-; b D - - Weekly need satisfied - -4. ✅ **Updated reset-to-first-launch.sh** - - Added missing transient files/directories - - Keeps docs/, inbox.org, .localrepo safe - - Ready for testing offline installs - -5. ✅ **Tested .localrepo offline install capability** - - Works perfectly for package.el packages - - Discovered 3 bugs during test (logged in inbox.org) - -*Bugs Discovered (all logged in inbox.org):* -1. **Chime throw/catch error** - High priority, your package - - Error: "(no-catch --cl-block-chime-check-- nil)" - - Fix: Change defun to cl-defun or add catch block - - Currently disabled to unblock startup - -2. **go-ts-mode-map keybinding error** - - Error: "void-variable go-ts-mode-map" - - Fix: Wrap in with-eval-after-load - -3. **Treesitter grammars not in .localrepo** (limitation documented) - - Expected behavior - treesit-auto downloads separately - -*Metrics Update:* -- Startup time: 6.19s → 4.16s (**2.03s improvement!**) -- Only 1.16s away from < 3s target! -- Active todos: ~23 items (hit < 20 goal when excluding tracking tasks!) -- Shipped items: 2 (network check, diff-buffer-with-file) - -*Key Files Modified:* -- [[file:../early-init.el][early-init.el]] - Network check removed, cj/use-online-repos simplified -- [[file:../modules/custom-buffer-file.el][custom-buffer-file.el]] - Added cj/diff-buffer-with-file -- [[file:../inbox.org][inbox.org]] - Fresh V2MOM-aligned todo list created -- [[file:../scripts/reset-to-first-launch.sh][reset-to-first-launch.sh]] - Updated with missing transient files -- [[file:someday-maybe.org][someday-maybe.org]] - Old todo.org archived here - -*Next Session (2025-11-01):* -**Two quick wins ready (15 min each):** -1. Fix cj/goto-git-gutter-diff-hunks (missing function) -2. Fix chime throw/catch bug (re-enable chime) - -**Then continue Method 1:** -- Optimize org-agenda (THE BOTTLENECK - 30s → <5s target) -- Fix org-noter (daily pain) -- Fix video/audio recording -- Fix mail attachments -- Fix grammar checker - -*Craig's Words:* -> "There will always be cool ideas out there to implement and they will always be a web search away." -Ruthless prioritization in action! Deleted research files, focused execution. - -*** 2025-10-30 Session 1 - V2MOM In Progress -*Time:* ~1 hour -*Status:* PAUSED - V2MOM 60% complete - -*What We Completed:* -1. ✅ Created docs/ directory structure -2. ✅ Created SESSION-HANDOFF-ACTIVE-PROJECT.org (this file) -3. ✅ Created emacs-config-v2mom.org -4. ✅ Created values-comparison.org (analysis doc) -5. ✅ Completed Vision (already existed, kept as-is) -6. ✅ Completed Values section (Intuitive, Fast, Simple) - - Intuitive: Muscle memory, mnemonics, which-key timing, "newspaper" code - - Fast: Startup < 2s, org-agenda is THE bottleneck, everything else acceptable - - Simple: Production software practices, simplicity produces reliability |