Skip to content

BACKLOG.md

Latest commit

 

History

History
204 lines (149 loc) · 10.4 KB

BACKLOG.md

File metadata and controls

204 lines (149 loc) · 10.4 KB

Reports Hub - Backlog

Open and deferred items. Completed items are in CHANGELOG.md.

Open: Medium Priority

NN1. Enhance lesson plan with LPTHW pedagogy elements — Done

Priority: Medium (learning experience, completeness) Source: Content review (Mar 25, neural-networks learning roadmap) + LPTHW audit hook Status: ✅ COMPLETED (Mar 25)

Delivered (205 lines, 2 commits: 0cef0ca, 10b75df):

  1. ✅ Learning objectives per phase (Phases 1-5)
  2. ✅ Per-phase glossary callout tables (plain-language definitions)
  3. ✅ "Try This" hands-on exercises per phase (paper sketch, NumPy loop, OTEL instrumentation, BertViz, quantization benchmark)
  4. ✅ Teach-Back prompts per phase (Phases 1, 2, 3, 5; Phase 4 capstone already had this)
  5. ✅ Production failure examples for Phases 2 and 3
  6. ✅ New "Common Training Pitfalls" section (loss spikes, dead neurons, observability gaps, OTEL anti-patterns)
  7. ✅ Teaching-background callouts ("As an educator, you know…") in each phase
  8. ✅ Role-based guidance was already present; left intact

[DONE]

NN2. Bridge neural network training to OTEL instrumentation (Highest-Leverage) — Done

Priority: HIGH (core competency gap, competitive advantage) Source: Content review (Mar 25) — identified as critical missing bridge Status: ✅ COMPLETED (Mar 25)

Deliverable: Internal guide "Instrumenting Neural Network Training for OTEL Observability" Save location: docs/neural-networks-otel-bridge.md

Delivered:

  1. ✅ Mapping table: Phase 2 training concepts → OTEL signals (8 rows: loss, gradients, backprop, neuron activation, learning rate, overfitting, timing, gradient flow)
  2. ✅ Concrete walkthrough: Instrumenting Karpathy's training loop
    • Before/after code snippets (Python + OpenTelemetry)
    • 3-line OTEL setup + 4 metric instruments
    • Instrumentation overhead noted as low (microseconds)
  3. ✅ Production patterns & alerting:
    • Alert thresholds table (warning/critical levels)
    • 2 example alert rules: overfitting detection, gradient explosion
  4. ✅ Failure case studies with instrumentation fixes:
    • Case 1: Silent NaN divergence (caught by gradient_nan_count + gradient_norm)
    • Case 2: Dead ReLU syndrome (caught by dead_neuron_ratio and layer-wise activation_mean)
    • Case 3: Overfitting creep (caught by loss_divergence ratio monitoring)
  5. ✅ Why this matters for new hires: Bridges theory → practice, competitive advantage, onboarding path
  6. ✅ Quick reference checklist: 8-point metric checklist for training scripts
  7. ✅ Extension to Phase 5: Fine-tuning, inference, quantization scenarios

Why this is highest-leverage:

  • ✓ No public article currently covers this specific bridge (verified via research)
  • ✓ Essential for OTEL-focused startup: ties training science to observability practice
  • ✓ Directly applicable to new hire's onboarding (cross-reference from lesson-plan Phase 2→3 bridge)
  • ✓ Reusable for internal training, documentation, and knowledge transfer

Effort: 4-5 hours (~3500 words, comprehensive with production patterns + 3 case studies)

[DONE]

NN3. URL verification for lesson-plan.md — Done

Priority: Critical (pre-distribution) Source: Quality audit (Mar 25) — hallucination flag on unverified URLs/dates

All 14 resource URLs verified via WebFetch (Mar 25, 2026):

  • ✅ All URLs active and accessible
  • ✅ All author attributions correct (3Blue1Brown, Karpathy, Lil'Log, OTEL, fast.ai, etc.)
  • ✅ Resource descriptions match actual content
  • ✅ Time estimates reasonable based on content scope
  • ✅ Future-dated resources (2025) confirmed authentic — we're now in March 2026, so these are contemporary

Hallucination risk resolved: The three "future-dated" resources (Rohan Paul explainability, LoRA/Quantization DEV article, OTEL AI Agent Observability post) that triggered warnings are legitimate 2025 publications. No fabrication; judge's caution was appropriate at evaluation time.

Lesson plan is CLEARED for distribution to new hire.

[DONE]

H11. Extract inline styles from 50 report files — Done (Phase 4)

Priority: Medium (code quality, maintainability) Source: CSS architecture audit (Mar 9) Phase 1 (Mar 9): Extended section h2, #toc h2 CSS selector; removed nav#toc and h2 redundant inline styles (29 instances); replaced color: white with color: var(--color-white) (11 files). ~290 inline styles remain (was ~367, excluding overflow-x:auto). Phase 2 (Mar 9): Added 24 utility classes to report-base.css. Replaced 150+ inline styles across 37 files. Extracted hardcoded SWOT hex colors to semantic CSS classes in edgar_nadyne files. Phase 3 (Mar 9): Added spacing tokens and SWOT utilities. Extracted 30 more inline styles from 6 edgar_nadyne files. Phase 4 (Mar 9): Added --color-swot-opportunity token to edgar-nadyne brand and .color-swot-opportunity utility (with dark mode override). Replaced #4a1a6b/#6B2D5B heading colors in artist profile files with .color-swot-opportunity/.text-primary. Replaced margin-top: 5px with .mt-xs in leora dashboards. Remaining: table column widths and overflow-x:auto wrappers intentionally left as-is (layout-critical, not extractable to shared utilities).

[DONE]

T1. Create content-translator skill — Done

Priority: Medium Source: Readability audit (Feb 13)

Built ~/.claude/skills/content-translator/SKILL.md — 5-phase workflow: source extraction, translation, localization, assembly, QA validation. Handles BCP-47 lang tags, source-tracking comments, skip-link translation, hub card and TRANSLATION_STATUS.md updates. Invoke via /content-translator <file.html> --lang <BCP-47>.

[DONE]

T5. Multi-language support beyond PT-BR — Done

Priority: Medium Source: Readability audit (Feb 13)

Added locale conventions table (19 BCP-47 tags: pt-BR, pt-PT, es, es-MX, es-ES, fr, fr-CA, de, he, it, ja, zh-CN, zh-TW, ko, ar, nl, pl, sv, tr) to content-translator SKILL.md. Covers date formats, decimal separators, currency symbols, quote marks, and RTL handling (ar, he). Replaces two ad-hoc inline bullet points with a structured, extensible reference table. Phase 3 localization table also updated with per-locale examples.

[DONE]

W1. Repomix granular compression config — track upstream

Priority: Medium Source: Fact-check audit (Feb 24), code-condense-whitepaper session Upstream: yamadashy/repomix #561, #516

Repomix --compress is currently all-or-nothing (output.compress: boolean). Granular controls (e.g., keep_signatures, keep_interfaces, keep_docstrings, per-directory compress patterns) are discussed upstream but not yet implemented. When upstream support lands, update code-condense-whitepaper/repomix-command-line-cheat-sheet.md with the actual config schema and remove the "Future Granularity" placeholder section.

Effort: Low (documentation update when upstream ships)

Deferred (P3-P4)

F6. Hub layout monotony

Priority: P4 Source: Bugfix plan item #12 (Feb 15)

index.html has 10 sections, most with only 1 card. Consider grouping related brands, visual hierarchy, or different card sizes.

Effort: High (design + implementation)

F7. CSS variable namespace consolidation — Done

Priority: P4 Source: Bugfix plan item #13 (Feb 15)

Added ASCII namespace reference table to theme.css documenting all four parallel namespaces (portal-base, report-base, competitor-base, holliday portal-layout). Added --color-* bridge aliases to report-base.css and competitor-base.css so cross-namespace utilities can reference a consistent --color-* convention without renaming any existing variables. Portal-base needs no bridge (already uses --color-* natively). --color-surface intentionally scoped to competitor-base only (card surface concept vs. page background in report-base). Zero regression risk — purely additive.

[DONE]

Reference

Completed Item Summary

See CHANGELOG.md for full details.

Session Items Completed
Feb 13 — Readability Audit C1-C3, H1-H4, M1-M6, L1-L4 (16 items)
Feb 13 — Translations T2, T3, T4 (3 items)
Feb 14 — DRY Review D1-D13 (13 items)
Feb 15 — Design Review S1 #1-#10 (10 items)
Feb 15 — Design Review S2 F1-F4 (4 items)
Feb 16 — OTEL Improvement Phase 2 F8-F12 (5 items)
Feb 18 — Accessibility Audit A1-A10 (10 items)
Mar 9 — CSS Refactor & Theme System A12, A13, H10, H12, H13, H14, W4 (7 items)
Mar 9 — Backlog Implementer A11, W2, W3, F5 (4 items)
Mar 9 — Backlog Implementer (2) H11-P3-A, H11-P3-B (2 items)
Mar 9 — Backlog Implementer (3) H11-P4, T1 (2 items)
Mar 9 — Backlog Implementer (4) T5 (1 item)
Mar 9 — Backlog Implementer (5) F7 (1 item)
Total 80 completed, 0 open

Scorecard (Phase 2 — Feb 16)

All 12 directories at grade A. Baseline → Phase 1 → Phase 2 progression:

Directory Baseline Phase 1 Phase 2 Dark Mode Mobile Tables A11y
balloon-collective/ D B A Pass Pass Pass Pass
capital_city/ A B A Pass Pass Pass Pass
edgar_nadyne/ D C A Pass Pass Pass Pass
holliday_lighting/ C C A Pass Pass Pass Pass
integrity-studio-ai/ D B A Pass Pass Pass Pass
leora_research/ D B A Pass Pass Pass Pass
ngo-market/ D D A Pass Pass Pass Pass
PerformanceTest/ D B A Pass Pass Pass Pass
skelton-woody/ D B A Pass Pass Pass Pass
sound-sight-tarot/ D B A Pass Pass Pass Pass
trp-austin/ D B A Pass Pass Pass Pass
zoukmx/ D B A Pass Pass Pass Pass

DRY Testing Checklist

  • All portal index pages render correctly
  • All report pages maintain brand theming
  • Competitor analysis tables display correctly
  • Dark mode switches properly on all pages
  • Responsive breakpoints work at 768px and 480px
  • Print styles maintain formatting
  • Leora referral form and dashboard still function
  • No visual regressions in badge styling