for corporate trainers & L&D experts · external sources + Karpathy wiki
Fast discovery
Deep synthesis
Orchestration / workflow
🐕⚡
Perplexity Pro
“Fast scout” – delivers cited answers in 2–4 min. Academic mode filters peer‑reviewed papers. Model Council runs GPT-4o/Claude/Gemini side‑by‑side.
🔗 cited web + scholar
⚡🔍
Gemini Flash / Search
Real‑time grounding, connects to MCP filesystem. Lightweight but powerful for scanning recent preprints & corporate repositories.
🌐 live data + wiki aware
🧩 Step 1: Use discovery agents to find knowledge gaps and fresh papers (last 7 days).
🐘📘
Gemini Deep Research Max
20–60 min autonomous research · reads 100+ sources · MMLU 95.1 (expert level). Outputs structured report with citations & contradiction analysis.
📑 deep synthesis
📦🧠
Karpathy‑style Wiki
Compound knowledge base: raw/ sources + entity pages + `index.md`. Never forgets past research. Auto‑linking & conflict detection. Grows every week.
🗂️ persistent memory
🧠 Step 2: Gemini writes deep report → Claude ingests everything into the wiki (auto‑updates index & log).
🔌🤝
MCP Bridge
Model Context Protocol connects Perplexity, Gemini and Claude to the same wiki folder. Agents share context, avoid double work, and query past findings natively.
🔄 unified memory
📝🧸
Claude Organizer
Turns wiki knowledge into training modules: slide decks, scenario‑based quizzes, SCORM ready. `/wiki:ingest` & `/wiki:query` slash commands.