Kaizen Craft

Nature as Algorithm

Network robustness, pollinator diversity, and learned embeddings of the waggle dance

Working notes on AI systems, safety research, labor, ethics, and governance policy

Setting up my Digital Brain Karpathy Wiki with RAG system and local DB, running on local models

How I turned my CarbonInterface learning journal into a locally-run RAG system: Postgres + pgvector, Ollama embeddings, a Karpathy-style regenerable wiki, and a daily brief that reads my calendar, email, and Scholar alerts for me.

Knowledge Management RAG growing

Jul 2026

My Google Interview, or Curriculum at Model Speed: A Multi-Agent System for Staying Ahead of AI

A design for a multi-agent pipeline that sources, vets, and scaffolds AI curriculum as fast as the field actually moves, plus metrics for measuring behavior change instead of course completions, plus a filterable table of the actual sources worth watching.

AI Agents & Tools Knowledge Management growing

Jul 2026

Matt Pocock's Skills, Actually Explained: A Critical Guide for Current Usage

A dig through Matt Pocock’s mattpocock/skills repo (the real SKILL.md files, not the SEO blogspam about it): what the skills actually do, what the community has genuinely pushed back on, and what’s worth stealing.

AI Agents & Tools Claude Code growing

Jul 2026

Literature Review 1.0: Red-Teaming, Quantization, and Multilingual Robustness

Phase 1 - Research Niche Selection & Literature Foundation.

Week 1 Literature Review growing

Jul 2026

When Search Goes Wrong

ICML 2026 — CREST-Search red-teams web-augmented LLMs by manipulating search queries to induce unsafe citations, not unsafe generation.

ICML Adversarial growing

Jul 2026

Gray-Box VLM Adversarial Alignment

ICML 2026 — SVD-structured adversarial attacks against gray-box vision-language models.

ICML Alignment growing

Jul 2026

See all field notes

Sketches

Experiments at different development stages

Digital Brain: Local RAG + Regenerable Wiki

Local-first personal knowledge stack: Obsidian vault → Postgres/pgvector → Ollama embeddings → entity graph → Karpathy-style regenerable wiki, with Claude Code over stdio MCP and a daily brief. Working for daily use; several surfaces still under active development.

RAG pgvector growing

Jul 16

CivicSpark AI

CivicSpark in context of peer civic-AI projects (Engagic, Open Council, Civic Band, View Royal, GRASP): common failure modes, and an iterative path from the Tulsa PoC toward grounded, item-level municipal systems.

civic tech RAG growing

Jul 16

Multi-Agent EDU Stack

Eight-stage multi-agent educational stack for sourcing, vetting, and scaffolding AI curriculum under high content decay — plus Level-3 behavior telemetry. Implementation repo in active development.

multi-agent systems curriculum design seeding

Jul 16

See all sketches

Map

Navigation across relational landscapes