Implementation ledger
7 layers fully implemented as RDF-Turtle rules; 3 partial (host harness); 1 external (model).
What this shows
The PDF is treated as an empirical scorecard: each stack layer is mapped to
real code under agent-rdf-memory, not a slide-only theory deck.
Source model by Brij Kishore Pandey; live mapping against the OpenLink
ai-agent-skills memory store.
Fail-closed policy
Identity, preferences gates, and evaluation scripts block non-compliant edits before they land.
Verified methodology
Links checked via GitHub Trees API before Linked Data generation — no guessed paths.
Slide arc (PDF pages)
- Title — Live RDF mapping: 11 layers, 140+ steps, 25 verified links.
- Iceberg — Reasoning model is ~10%; engineering is the other 90%.
- Ledger — Layer-by-layer status against code artifacts.
- Context & Memory — ontology.ttl funnel; sessions / entities / preferences.
- Skills & Orchestration — howto/ modules meshed with AGENTS.md protocol.
- Fail closed — core.ttl identity, preferences gates, evaluation scripts.
- Boundary map — in-repo contracts vs host harness, tools, Virtuoso, LLM.
- Hub-and-spoke — preferences.ttl hub; companion howto spokes; private overlay.
- Session-start protocol — mandatory retrieval path with SPARQL fallback.
- Verification chain — five-step methodology of truth.
- Blueprint — shareable, verifiable operating environment.
Production facts
Duration
2:13
Resolution
1440×900
Voice
onyx TTS
PDF pages
11
Recording pipeline
PDF page renders → HTML scene deck
(presentation-scenes-recording-deck.html) →
shot-scraper video silent record → OpenAI TTS MP3 (onyx) →
ffmpeg mux to final MP4. Bundle lives under the Grok
screencasts/ output root (not webpages/).
Narration
Full voice-over script used for the onyx TTS track.
Show narration.txt
Loading…