Wiki Index

Catalog of all knowledge pages. Maintained by Claude. Last updated: 2026-04-24 (Alex Krantz / OpenClaw architecture deep dive).

Concepts

Ideas, frameworks, mental models

• llm-wiki-pattern — LLM-maintained wiki where the LLM writes all pages
• coding-agent — Six-component harness (context, cache, tools, compaction, memory, subagents) that wraps an LLM for coding
• thin-harness-fat-skills — Keep the harness minimal; encode domain intelligence in reusable markdown skill files
• latent-vs-deterministic — The judgment/trust line every agent step sits on; artifact-dependency rule; three-column ownership
• claude-code-skills — Nine skill types, writing tips, and distribution patterns for Claude Code skills
• dimensional-modeling — Kimball’s methodology for DWH design using fact + dimension tables
• star-schema — Central fact table surrounded by denormalized dimension tables
• fact-table-types — Transactional, periodic snapshot, accumulating snapshot, factless, aggregate
• slowly-changing-dimensions — SCD types 0–7 for handling dimension attribute history
• etl-architecture — Extract-transform-load pipeline that feeds a data warehouse
• medallion-architecture — Bronze/Silver/Gold layered pipeline combining Kimball modeling with a semantic layer
• semantic-layer — Governed metric definitions as first-class database objects above the dimensional model
• agent-sandbox — Per-user isolated execution environments with S3-backed mounts and AWS ABAC credentials
• context-normalization — Converting heterogeneous source data into uniform markdown/CSV/JSON for agent reasoning
• s3-first-architecture — S3 as durable source of truth with PostgreSQL as a query-optimized read layer
• agent-evaluation — Domain-specific eval suites and eval-driven development for production AI agents
• model-context-protocol — Open standard giving AI models universal access to external tools, data, and services
• agentic-engineering — Four-layer framework (Configuration, Capability, Orchestration, Guardrails) for reliable coding agents
• prototype-and-prune — AI-era product development replacing stage-gate planning with rapid parallel experimentation
• agent-memory — Three-layer memory model: MEMORY.md, episodic-memory (vectors), QMD (on-device search)
• product-operating-model — Marty Cagan’s framework: empowered teams, outcomes over outputs, continuous discovery
• design-taste-craft — Dylan Field’s three-part distinction: taste (preferences), craft (micro+macro), point of view (unique insight)
• barrels-and-ammunition — Rabois’s hiring framework: barrels own outcomes end-to-end; ratio of barrels determines parallel initiative capacity
• agent-first-software — Software built for agent consumption: 100–1000x agent volume, semantic selection, enterprise diffusion gap
• auto-research — Autonomous agent loop: objective + metric + boundaries; program.MD; SETI@home distributed pattern
• product-trio-agentic-era — How PM/Designer/Engineer roles evolve as agents absorb execution; small team and startup contexts
• dark-code — Emergent, unattributable production behavior from agent runtime decisions; the governance failure mode
• context-engineering — 20+ year discipline of bridging human intent and machine understanding; four-era evolution model
• agentic-system-failure-modes — Six production failure patterns that silently destroy agents; diagnostic framework from 2 years in production
• ai-agent-observability — OTel GenAI conventions, trace primitives, tool landscape; direct response to dark code accountability gap
• agent-human-collaboration — Trust/control axes, work-as-DAG, high-bandwidth artifacts over chat for complex agent work

People

Notable people worth tracking

• andrej-karpathy — AI researcher, educator, OpenAI co-founder; auto research, macro actions, loopy era
• sebastian-raschka — ML researcher, author of Build a LLM From Scratch, writes on agents and deep learning
• garry-tan — YC president; thin harness, fat skills architecture pattern
• ralph-kimball — Originator of dimensional modeling and the Kimball DWH methodology
• bill-inmon — “Father of the Data Warehouse”; coined the term and top-down methodology
• nicolas-bustamante — Founder of Fintool; builder of AI agents for professional investors in financial services
• julie-zhuo — Former VP Design at Facebook; author of The Making of a Manager; writes The Looking Glass
• teresa-torres — Product discovery coach; author of Continuous Discovery Habits; empirical AI product team researcher
• paul-hoekstra — Data engineer; writes Paul’s Pipeline; Agentic Engineering four-layer framework
• marty-cagan — SVPG founder; product operating model; empowered teams framework
• dylan-field — Co-founder/CEO of Figma; “design is the new code”; taste/craft/POV framework
• keith-rabois — Partner at Khosla Ventures; barrels/ammunition, anti-customer-feedback, undiscovered talent
• aaron-levie — Co-founder/CEO of Box; agent-first software design; SaaS-as-agent-backend
• marc-andreessen — Co-creator of Mosaic/Netscape; a16z; Unix agent architecture; death of UI
• claire-vo — Founder of ChatPRD; 9-agent personal OpenClaw stack; AI-native operations practitioner
• felix-lee — CEO of ADPList; ships production apps without Figma; taste codification practitioner
• sarah-guo — Conviction VC founder; coined “dark code”; cross-portfolio AI risk pattern recognition
• jensen-huang — Nvidia co-founder/CEO; electrons→tokens; five-layer AI stack; CUDA ecosystem; supply chain moat
• owen-jennings — Block business lead; 40% RIF; Goose harness; generative UI; proprietary signal moat thesis
• jacob-lauritzen — CTO of Legora; trust/control framework; high-bandwidth artifacts for agent-human collaboration
• tuomas-artman — CTO of Linear; Quality Wednesdays, zero bug policy, AI taste gap, product engineer convergence
• nikhyl-singhal — Skip community founder; information-mover PM is dying; 30k shed/8k hire; logo depreciation; PM dandelion seeds
• alex-krantz — UC Berkeley PhD; OpenClaw architectural analysis; loopiness model; sessions-as-processes; design > implementation

Projects

Ongoing or tracked projects

• active-projects-2026 — Snapshot of active development projects and Claude Code usage patterns as of April 2026

Resources

Books, papers, tools, courses

• obsidian — Local-first markdown editor with graph view and backlinks
• mcp-servers — Curated reference of 40 tested MCP servers organized by category
• openclaw — Open-source personal AI agent platform: gateway, multi-agent, cron, channels
• figma — Figma MCP workflows, design-to-code, code-to-Figma, Figma Make vs Claude Code

Topics

Subject areas and domains

• data-warehouse — Subject-oriented, integrated, time-variant, non-volatile analytics data store
• ai-agents — LLMs wrapped with tools, memory, and skills to autonomously complete multi-step tasks
• product-development — Practices and team structures for taking ideas to shipped product in the AI era