Show HN: 10x better performance from the Coding Harnesses with LLM-wiki
4 hours ago
- #knowledge-management
- #markdown-wiki
- #multi-agent-research
- Parallel multi-agent research system with 5-10 agents searching academic, technical, applied, news, and contrarian angles
- Rounds-based research that drills into gaps found each round, with minimum time options (e.g., 2h)
- Source ingestion from URLs, files, PDFs, inbox drops, Git repos, MediaWiki dumps, and archives
- Wiki compilation with cross-referenced articles, confidence scores, and session memory
- Multiple installation modes: Claude Code plugin, OpenAI Codex plugin, OpenCode instruction file, Pi for local models, and portable AGENTS.md
- Topic isolation with separate wikis, archive capability, and Obsidian compatibility
- Operational session memory with redacted checkpoints, digests, and feedback candidates stored in .sessions/
- Output generation for reports, slide decks, study guides, playbooks, and implementation plans
- Trust audits that trace evidence chains and perform fresh research when needed
- Inventory tracking for items, source candidates, corpora, entities, and datasets with manifests