📄 2026-04-27.md 15,876 bytes Apr 27, 2026 📋 Raw

2026-04-27 — Daily Blog Post: Scope Creep + Inline Wired SVGs

Blog Post — Scope Creep

  • Drafted and published: "Project Icarus and the Scope Creep That Nobody Warned Me About"
  • Category: openclaw
  • Tags: agents, openclaw, project-management, scope-creep, lessons-learned
  • Cover image: scope-creep-wired.svg (Wired-style hero with expanding concentric rings visualizing scope creep)
  • Live: https://hoffdesk.com/blog/project-icarus-scope-creep/

Wired-style Hero SVG

  • Created scope-creep-wired.svg with visual metaphor:
  • Core circle (cyan) = original V1.0 scope
  • Expanding rings (amber → red) = scope creep layers
  • Animated pulse on outer ring = ongoing expansion
  • Diagonal slash + glitch bands = Wired aesthetic
  • Stats badge: "⚠ 90 MIN → 3 HOURS → SHIPPED"
  • 5.9KB pure SVG, duotone gradients, grain noise overlay

Post Summary

  • Theme: Managing scope creep when your builders are AI agents
  • Hook: Recipe toggle feature that ballooned into a full system
  • Key insight: Agents optimize for completeness, not necessity
  • Three fixes: acceptance criteria first, hard timeboxes, ship before polish
  • Closing: Aundrea as the real acceptance criteria

Inline Wired-Style SVGs Added (2026-04-27)

  • Created 5 inline SVG illustrations in /blog/static/images/inline/ for section-level art in post bodies:
  • dns-broken-node-wired.svg — 3-node network topology with broken upstream (DNS post)
  • three-agents-wired.svg — Three agent boxes (Wadsworth/Daedalus/Socrates) with arrows (Board post)
  • full-stack-diagram-wired.svg — Layered sovereign stack (JS-free post)
  • learning-curve-chart-wired.svg — 4-phase curve chart with animated glow (Learning Curve post)
  • content-pipeline-wired.svg — Horizontal pipeline flow: markdown → API → Jinja2 → HTML (Blog post)
  • Added .blog-inline-art CSS class — centered, rounded, dark-bg, responsive (max 700px, full-width mobile)
  • Bumped CSS version to ?v=5 for cache-busting
  • Injected one SVG per post at midpoint section break
  • Fixed: /blog/ legacy URL routing (301 redirects to /api/blog/article/{slug} etc.)

2026-04-27 — Daily Blog Post: Scope Creep + Inline Wired SVGs

Blog Post — Scope Creep

  • Drafted and published: "Project Icarus and the Scope Creep That Nobody Warned Me About"
  • Category: openclaw
  • Tags: agents, openclaw, project-management, scope-creep, lessons-learned
  • Cover image: scope-creep-wired.svg (Wired-style hero with expanding concentric rings visualizing scope creep)
  • Live: https://hoffdesk.com/blog/project-icarus-scope-creep/

Wired-style Hero SVG

  • Created scope-creep-wired.svg with visual metaphor:
  • Core circle (cyan) = original V1.0 scope
  • Expanding rings (amber → red) = scope creep layers
  • Animated pulse on outer ring = ongoing expansion
  • Diagonal slash + glitch bands = Wired aesthetic
  • Stats badge: "⚠ 90 MIN → 3 HOURS → SHIPPED"
  • 5.9KB pure SVG, duotone gradients, grain noise overlay

Post Summary

  • Theme: Managing scope creep when your builders are AI agents
  • Hook: Recipe toggle feature that ballooned into a full system
  • Key insight: Agents optimize for completeness, not necessity
  • Three fixes: acceptance criteria first, hard timeboxes, ship before polish
  • Closing: Aundrea as the real acceptance criteria

Inline Wired-Style SVGs Added (2026-04-27)

  • Created 5 inline SVG illustrations in /blog/static/images/inline/ for section-level art in post bodies:
  • dns-broken-node-wired.svg — 3-node network topology with broken upstream (DNS post)
  • three-agents-wired.svg — Three agent boxes (Wadsworth/Daedalus/Socrates) with arrows (Board post)
  • full-stack-diagram-wired.svg — Layered sovereign stack (JS-free post)
  • learning-curve-chart-wired.svg — 4-phase curve chart with animated glow (Learning Curve post)
  • content-pipeline-wired.svg — Horizontal pipeline flow: markdown → API → Jinja2 → HTML (Blog post)
  • Added .blog-inline-art CSS class — centered, rounded, dark-bg, responsive (max 700px, full-width mobile)
  • Bumped CSS version to ?v=5 for cache-busting
  • Injected one SVG per post at midpoint section break
  • Fixed: /blog/ legacy URL routing (301 redirects to /api/blog/article/{slug} etc.)

Inline SVG Color Unification (2026-04-27, post-flush)

  • Rebuilt all 5 inline SVGs with exact 7-color hero palette:
  • Indigo #6366f1 — primary elements (Pi-hole node, agent cards, stack layers)
  • Cyan #00e5ff — secondary elements (router, Daedalus, output glow)
  • Amber #f59e0b — warnings/agents (Socrates, frustration phase marker)
  • Red #ef4444 — failures (FAILED node, broken link X, glitch bands)
  • Magenta #ff00e5 — accents (pipeline Jinja2 box, denial phase marker)
  • Purple #7c3aed — connectors/transitions (arrows, FastAPI layer, stack bracket)
  • Background #0a0a12 — dark canvas
  • Added consistent visual flourishes: grain noise overlays, animated pulse dots, dashed stroke animations, glitch bands
  • Removed orphan original SVGs (board-architecture.svg, cost-comparison.svg, dns-spof.svg, full-stack.svg, learning-curve.svg) — no longer referenced by any template
  • All 5 blog posts verified rendering inline SVGs correctly on public site

Family Brain — Live Network Visualization (2026-04-27)

Matt asked for the Family Brain SVG to be live — fed by real data, auto-scaling, dynamic.

What was built:

  1. Stats endpoint GET /api/system/brain/stats — queries 6 data stores live:
    - Blog: SQLite post counts (9 total, 7 published)
    - Ontology: Entities & relations (10 entities, 9 relations)
    - Memory: ChromaDB chunk count + file size (~53MB across 3 agents, 795 chunks)
    - Calendar: Radicale systemctl status ("active")
    - LLM: Ollama model count (9 models)
    - Recipes: Ontology recipe-type entities (0 saved)
  2. SVG generator GET /api/system/brain/svg — renders full network SVG via Jinja2 template
    - Auto-scaling: Blog node grows with posts, LLM with models, Ontology with entities
    - Animated: 6 data packet dots traveling along connections, pulsing node glows
    - Live subtitle: e.g. "6 DATA STORES · 52896KB VECTORS · 7 BLOG POSTS · 9 LLM MODELS"
  3. Template: blog/templates/brain_svg.svg.j2 — Jinja2 SVG with Wired aesthetic
  4. Preview page: blog/static/brain-live.html — HTMX auto-refresh every 60s
  5. Spec: shared/project-docs/family-brain-spec.md — documented for Socrates

Files changed:

  • family/brain.py (NEW) — stats + SVG generator
  • main_v2.py — brain router mounted on blog_app + family_app + dev_app
  • shared/session_auth.py — added /api/system/ to EXEMPT_PREFIXES
  • blog/templates/brain_svg.svg.j2 (NEW) — Jinja2 SVG template
  • blog/static/brain-live.html (NEW) — HTMX preview page
  • shared/project-docs/family-brain-spec.md (NEW) — spec & docs

Live URLs:

  • SVG: https://hoffdesk.com/api/system/brain/svg
  • Stats: https://hoffdesk.com/api/system/brain/stats
  • Preview: https://hoffdesk.com/blog/static/brain-live.html

Next:

  • Phase 2: Socrates adds query_count tracking, last_activity for all stores, cross-connections
  • Phase 3: Clickable nodes with detail overlays, animated transitions

Second Blog Post: Waiting for UAT (2026-04-27)

  • Drafted and published: "Waiting for the User: Why UAT Is the Hardest Part of Building for Your Family"
  • Category: openclaw
  • Tags: adoption, lessons-learned, openclaw, project-management, uat
  • Cover image: waiting-for-uat-wired.svg (hourglass with family silhouette, Wired aesthetic)
  • Draft source: /home/hoffmann_admin/.openclaw/shared/project-docs/blog/draft-waiting-for-uat.md
  • Created via PATCH on /api/blog/admin/posts/{slug}
  • Published at 2026-04-27T17:11:32.913721Z
  • Live: https://hoffdesk.com/blog/waiting-for-the-user/
  • Theme: The agony of waiting for real users (family) to test your app, edge cases discovered in the gap between "works on my machine" and "works for Aundrea at 7 AM"

Inline SVG Color Unification (2026-04-27, post-flush)

  • Rebuilt all 5 inline SVGs with exact 7-color hero palette:
  • Indigo #6366f1 — primary elements (Pi-hole node, agent cards, stack layers)
  • Cyan #00e5ff — secondary elements (router, Daedalus, output glow)
  • Amber #f59e0b — warnings/agents (Socrates, frustration phase marker)
  • Red #ef4444 — failures (FAILED node, broken link X, glitch bands)
  • Magenta #ff00e5 — accents (pipeline Jinja2 box, denial phase marker)
  • Purple #7c3aed — connectors/transitions (arrows, FastAPI layer, stack bracket)
  • Background #0a0a12 — dark canvas
  • Added consistent visual flourishes: grain noise overlays, animated pulse dots, dashed stroke animations, glitch bands
  • Removed orphan original SVGs (board-architecture.svg, cost-comparison.svg, dns-spof.svg, full-stack.svg, learning-curve.svg) — no longer referenced by any template
  • All 5 blog posts verified rendering inline SVGs correctly on public site

Family Brain — Live Network Visualization (2026-04-27 evening)

Matt asked for the Family Brain SVG to be live — fed by real data, auto-scaling, dynamic with Wired aesthetic.

What was built:

  1. Stats endpoint GET /api/system/brain/stats — queries 6 data stores live:
    - Blog: SQLite post counts (9 total, 7 published)
    - Ontology: Entities & relations (10 entities, 9 relations)
    - Memory: ChromaDB chunk count + file size (~53MB across all 3 agents, 795 chunks)
    - Calendar: Radicale status from systemctl (active, 0 collections)
    - LLM: Ollama model count (9 models)
    - Recipes: Ontology recipe-type entities (0 saved)
  2. SVG generator GET /api/system/brain/svg — renders full network SVG via Jinja2 template
    - Auto-scales: Blog node radius with post count, LLM with models, Ontology/Memory with record counts
    - Animated: 6 data packet dots traveling along connection paths at varying speeds
    - Live subtitle line: e.g. "6 DATA STORES · 52896KB VECTORS · 7 BLOG POSTS · 9 LLM MODELS"
    - Wired aesthetic throughout: grain noise filter, duotone gradients, glitch bands, dark backdrop
  3. Template: blog/templates/brain_svg.svg.j2 — Jinja2 SVG with embedded CSS animations
  4. Preview page: blog/static/brain-live.html — uses HTMX to auto-refresh SVG every 60s
  5. Spec: shared/project-docs/family-brain-spec.md — documented for Socrates

Files created/changed:

  • family/brain.py (NEW) — stats collection + SVG generator logic
  • main_v2.py — brain router mounted on blog_app (public) + family_app (auth-exempt) + dev_app
  • shared/session_auth.py — added /api/system/ to EXEMPT_PREFIXES
  • blog/templates/brain_svg.svg.j2 (NEW) — Jinja2 SVG template
  • blog/static/brain-live.html (NEW) — HTMX preview page
  • shared/project-docs/family-brain-spec.md (NEW) — full spec doc

Live URLs:

  • SVG: https://hoffdesk.com/api/system/brain/svg
  • Stats: https://hoffdesk.com/api/system/brain/stats
  • Preview: https://hoffdesk.com/blog/static/brain-live.html

Node layout (orbital):

  • CENTER: BEELINK (r=58, cyan)
  • TOP-LEFT: CALENDAR (r=38, indigo)
  • TOP-RIGHT: MEMORY (r=48, magenta)
  • LEFT: BLOG (r=45, purple)
  • RIGHT: LLM (r=39, amber)
  • BOTTOM-LEFT: RECIPES (r=34, cyan, grayed — 0 records)
  • BOTTOM-RIGHT: ONTOLOGY (r=38, red)

Phase 2 backlog (needs Socrates):

  • Add query_count tracking per service
  • Add last_activity timestamps for Radicale, Ollama
  • Add health checks per store (port/process ping)
  • Cross-connections between satellites (e.g. Blog → Memory for vector search)

Key Decisions (2026-04-27)

  • Use PATCH (not PUT) for blog admin post updates
  • Jinja2 template path for SVG must resolve to /home/hoffmann_admin/hoffdesk/blog/templates/
  • Brain endpoint publicly accessible (no auth) since it's informational, not administrative
  • HTMX polling on dashboard every 60s for brain refresh — not faster

Family Brain — SVG Rendering Bug Fixed (2026-04-27 late evening)

Matt reported: "blue box with question mark" when viewing the brain preview page.

Root cause: Raw & character in SVG subtitle text: "Entities & Relations" — XML parsers reject unescaped ampersands.

Fixes applied:
1. Added <?xml version="1.0" encoding="UTF-8"?> declaration to SVG output (Jinja2 template)
2. Escaped subtitle text in brain.py: "Entities &amp; Relations"
3. Added _escape() helper to brain.py for all dynamic text fields
4. Verified with browserless screenshot: renders clean with all 6 satellites visible

URLs confirmed working:
- https://hoffdesk.com/api/system/brain/svg (now valid SVG)
- https://hoffdesk.com/blog/static/brain-live.html (renders correctly)

Files modified:
- family/brain.py — added escape helper, fixed subtitle, added XML declaration
- blog/templates/brain_svg.svg.j2 — no change (escape happens in generator)

Research: Cognitive Load & Reminder Optimization for Family Brief

Matt asked for psychology research on reminders, scheduling, and brief structure optimization for family assistant and Icarus long-term.

Key Findings from Literature

1. Target-Action Reminders (Peper et al. 2022)
- Reminders must include both cue AND action to be effective
- "Pick up Sullivan" = weak. "Exit by 3:15 → drive to school → wait in car line" = strong
- Application: Every brief reminder is a mini-script, not a label

2. Hierarchical Progressive Disclosure
- Matches how prospective memory retrieves: gist first, detail on demand
- Four levels align with cognitive load stages:
- Month: Pattern recognition (1-2 sec glance)
- Week: Sequencing, conflicts (3-5 sec)
- Day: Execution details (5-10 sec)
- Hour: Full context + fallback (10-15 sec)

3. Implementation Intentions (Gollwitzer)
- Formula: When [cue], then [action]
- Transforms time-based intentions into cue-driven automatic behaviors
- Day view should frame as chain of when/then transitions, not event list

4. Cost Avoidance
- People use reminders when forgetting penalty > effort of reminder
- Brief should surface high-cost items first (not urgency but forgetting penalty)
- For 7 AM Aundrea: school lunch > forgotten = hangry kid + guilt + scramble

5. Ambient Visibility > Push Notifications
- Environmental cues outperform push alerts (interruption vs recognition cost)
- Brain habituates to push alerts in ~3 days (alert fatigue)
- UAT feedback about "more passive visibility" is validated by research

Strategies for Brief Structure

Strategy Application
Conflict Nag Window Visual density comparison, not just overlap alert
Evening Glance Mode 10-second "what's tomorrow" for bedtime encoding
Decision Boundaries Organize by natural transitions (pre-kids-up, lunch, pickup) not clock time
Forgetting Budget Show commitment count vs threshold: "14 active, budget ~8"
Anchor Event One anchor per day structures everything around it

Proposed Brief Hierarchy

ONE MONTH (28 dots, pure pattern)
  • Busy days highlighted (3+ commitments)
  • Anchor events marked
  • No text — visual density only

ONE WEEK (7 cards, horizontal)
  • Each: anchor event + commitment count
  • Conflicts = red dots
  • Label: Light/Medium/Heavy per day

ONE DAY (morning brief)
  • When/then chain framing
  • School lunch, dinner plan first
  • Commitment gauge visible

ONE HOUR (on tap detail)
  • Target + action + materials + fallback
  • Implementation intention format

UAT insight validated: Month overview + grounded information + passive visibility are the three pillars. The four-level hierarchy delivers all three.

Next: Spec data model for four-level brief structure, draft Month/Week/Day/Hour templates.

← Back