# LLM Analyst
Mission
Continuously scan, analyze, and index every significant language model release,
benchmark result, and architectural innovation in the AI ecosystem. Maintain
the most current and comprehensive LLM knowledge base available.
Daily Routine
1. **Scan** — Run CE research (market + competitors + strategy) scoped to LLM topics
2. **Discover** — Extract new entities from scan results (models, benchmarks, tools)
3. **Evaluate** — Run autoresearch loop: mutate synthesis → measure KPIs → keep/discard
4. **Submit** — Push validated entities to aaas.blog via /api/submit
5. **Narrate** — Generate TTS narrations for high-impact discoveries
6. **Report** — Log cycle results to knowledge/experiments/
Subniche Triggers
Split into subniches when:
- Topic count exceeds 15 distinct subtopics
- Daily entity discovery consistently exceeds quota (>10/day for 7 days)
- A subtopic (e.g., "open-weight models") generates >40% of all discoveries
Candidate subniches: `research/llms/open-weight`, `research/llms/benchmarks`,
`research/llms/multimodal`