Wei Cross Research
by @mikehankk
Cross-validate research answers by querying multiple LLMs in parallel with judge-based synthesis. Reduces hallucination and surfaces model disagreements for...
Example 1: Financial Query
Query: "็พ่ๅจ2026ๅนดไผ้ๆฏๅ๏ผ"
Selection process:
1. Keywords: ็พ่ๅจ, ้ๆฏ โ queryType: financial
2. Pass to skill: { "query": "...", "queryType": "financial" }
3. Skill looks up: config.json โ routing.financial.models
4. Skill selects: First 2 models from the routing config
5. Judge: Uses judge_financial.txt (Bull/Bear/Base Case analysis)
Example 2: Technical Query
Query: "How do I implement a distributed transaction?"
Selection process:
1. Keywords: implement, distributed โ queryType: technical
2. Pass to skill: { "query": "...", "queryType": "technical" }
3. Skill looks up: config.json โ routing.technical.models
4. Skill selects: Models configured for technical queries
Example 3: Social Query
Query: "What are people saying about SpaceX on Twitter?"
Selection process:
1. Keywords: Twitter, saying โ queryType: social
2. Pass to skill: { "query": "...", "queryType": "social" }
3. Skill looks up: config.json โ routing.social.models
- Note: grok-4.1 has social, sentiment, trending roles + X data access
clawhub install wei-cross-research