A personal knowledge base following the Zettelkasten methodology, maintained by LLMs using the LLM Wiki pattern.
Interlinked notes organized by topic. Browse in Obsidian for graph view and backlinks, or read the full index.
2026-04-20GeckoTerminal API evaluation - Keyless H1/H4 OHLCV for DEX pools; 14/15 ADOPT; same data as $129/mo CoinGecko Pro2026-04-19Static outbound IP solutions for crypto trading bots, ten options compared - 10 options across 5 categories; every commercial static-IP service re-bundles "rent a VPS" with markup2026-04-19Why rotating ISP IPs break Binance API keys, and how to fix it with WireGuard - Cheap VPS + WireGuard beats every bundled static-IP product for exchange whitelisting2026-04-19deepagents vs OpenClaw vs Hermes: category positioning - Library vs runtime distinction; the three are not peers and stack rather than compete2026-04-19AI tooling stack synthesis April 2026 - Synthesis: 3 layers wired through one rubric; growth and adoption-readiness are inversely correlated2026-04-19LLM agent memory synthesis April 2026 - Synthesis: 5-stage pipeline + 3 battlegrounds + harness hooks form one stack with a broken evaluation floor2026-04-19OpenBB evaluation - Python-first financial SDK; 11/15 BOOKMARK; crypto value thin, TradFi inflection point2026-04-19OSS trading stack survey, April 2026 - 3-category synthesis; Freqtrade + VectorBT canonical for semi-pro crypto; ai-hedge-fund as first agentic pilot2026-04-19FinceptTerminal evaluation - AGPL-3 Qt6 Bloomberg-alternative; 10/15 BOOKMARK; §13 blocks integration into any network-facing trading stack2026-04-19How LLM agents do web research: the ReAct loop - Agent research is a ReAct loop, not an algorithm; biggest failure is under-weighting Reddit/HN/Twitter2026-04-18Hermes Agent comprehensive briefing April 2026 - Nous Research's self-hosted agent with auto-generated skills; 0 to 95.6K stars in seven weeks2026-04-18Hermes vs OpenClaw competitive scene - OpenClaw wins metrics, Hermes wins narrative; equilibrium is to run both and bridge via ACP2026-04-18Why developers migrate to Hermes, real vs hype - Push factor (CVEs + subscription cliff) beats pull factor; steal the auto-skill pattern2026-04-18OpenClaw virtual company pattern - The "CEO/CTO/PM" multi-agent idiom is a convention, not a feature; six failure modes2026-04-18OpenClaw multi-persona dev team setup playbook - End-to-end config for a Telegram-led PM/Engineer/QA team on Docker2026-04-18TurboQuant KV cache compression - Random rotation + two-stage quantizer cuts KV cache to 3-4 bits with unbiased inner products
Obsidian Clipper (raw) ──┐
Claude Code (refined) ───┤──► this repo ──► Obsidian reads/lints/links
Claude AI skill (refined)┘
- Ingest: new knowledge arrives from web clips, coding sessions, or research conversations
- Compile: the LLM checks for overlaps, flags contradictions, updates cross-references and synthesis pages
- Query: ask questions against the compiled wiki; good answers get filed back as new notes
- Lint: periodic health checks for orphans, broken links, stale claims
The human thinks and curates. The LLM handles the bookkeeping.
Type these to Claude in this repo (exact wording doesn't matter, intent does):
| Command | What it does |
|---|---|
process inbox |
Refine raw _inbox/ captures into the right folder with links |
capture this as a note: ... |
Save a learning moment from the current chat |
compile recent commits / re-ingest |
Run compilation on notes pushed directly to GitHub (backlinks, index, README, log) |
answer from the wiki: ... |
Synthesize across notes, cite with [[wikilinks]] |
file that answer as a wiki page |
Commit a synthesis from chat as a new note |
reorganize the wiki |
Clean up orphans, folders, backlinks |
lint the wiki |
Health check: orphans, broken links, raw stragglers |
Full reference: commands cheatsheet in the usage guide.
| Folder | Domain |
|---|---|
| engineering/ | Software engineering practices, principles, languages |
| life/ | Life philosophy, habits, mindset, career wisdom |
| leadership/ | Management, negotiation, business leadership |
| cs/ | Computer science fundamentals |
| dwarves-kit/ | Dwarves Kit architecture and design |
| hiring/ | Candidate assessment and hiring |
| ai-tooling/ | AI developer tools and evaluations |
| ai/ | AI concepts, memory systems, agent patterns |
| diaspora/ | Vietnamese and Asian diaspora analysis |
| history/ | History, civilizations, geopolitical patterns |
| claude-code/ | Claude Code hooks, skills, workflows |
| crypto/ | Cryptocurrency, blockchain, DeFi, tokenomics |
| wealth/ | Trust-building, business relationships |
| geopolitics/ | Oil crises, government structures |
| startup/ | Startup strategy, product focus |
| health/ | Wellness, nutrition |
| investing/ | Compound interest, startup investing |
| mcp/ | Model Context Protocol |
| pkm/ | Personal knowledge management |
| devtools/ | Developer tools and config |
| patterns/ | Software patterns and anti-patterns |
| finance/ | Bond markets, capital structure, compounding knowledge |
| finance-tooling/ | Financial tool evaluations: terminals, data providers, broker platforms, frameworks |
| youtube/ | YouTube tooling |
| zed/ | Zed editor agent rules and configuration |
- Full note index - catalog of all notes with one-line summaries
- Usage guide - how to add notes, use Obsidian, work with Claude
- Architecture - system design, folder conventions, operations model
- Requirements - feature tracker, design principles, scaling triggers
- Changelog - project decisions log
- Operations log - chronological record of ingests, queries, lints