Projects

The Map

Every project serves one of two focuses. Most serve both. Together they form a self-improving stack where each layer watches, measures, and improves the one below it.

Learning Experiments

Hypothesis-driven work in unfamiliar domains. Every project decision becomes a testable claim with a verifiable outcome. 25+ experiments tracked, some confirmed, some refuted.

7 projects

Self-Improving Agents

Systems with feedback loops that make themselves better. Self-improving skills, self-healing tests, self-documenting history, self-managing context, self-improving data models, self-orchestrating compute.

4 projects

The Stack

Self-improving agent stack: each layer watches the one below

Each layer monitors the one below. BloomNet watches Dakka. Dakka orchestrates Claude Code. Claude Code uses BloomNet for context and memory. BloomNet triggers Skill Sync. The vault feeds experiments. Experiments drive the ratchet.

Core Projects

Oil Model Preferences

Commodity Futures · Learning Experiments

Monte Carlo simulation for CL futures with auto-ratcheting parameters against prediction markets.

+2.16pp Brier vs Polymarket GitHub →

Self-improving data models: 7 MC params auto-ratchet against live Polymarket prices

Jobs-Apply Preferences

Career Automation · Both Focuses

6-channel job automation with CDP browser control, behavioral adaptation, and J-Score v2 matching.

6 channels, 333+ applications GitHub →

Self-improving scoring: J-Score v2 three-layer fusion learns from application outcomes

Dakka Persistence

Agent Orchestration · Self-Improving Agents

Parallel Claude Code orchestrator with live API rate limit monitoring, spawn/kill UI, and Tauri v2 desktop.

7.8K lines Rust, Tauri v2 GitHub →

Self-orchestrating compute: parallel agents with live resource monitoring and self-spawning

BloomNet Memory

Developer Intelligence · Self-Improving Agents

Unified developer intelligence system: usage analytics, dynamic session context curation with adaptive forgetting, and multi-channel agent notifications.

Analytics + context curation + alerts GitHub →

Self-managing context + self-monitoring usage: scores, loads, and forgets context dynamically; visualizes the system watching itself; routes agent alerts across channels

RedCorsair Preferences

Brand System · Learning Experiments

Brand system with R-generated design tokens, Paul Tol colorblind-safe palette, publication-quality chart themes.

Slate + Journal design tokens GitHub →
Quick-Fin Preferences

Financial Tooling · Learning Experiments

MCP server for QuickBooks Online. 30-iteration audit across CPA, DE, and CFO personas.

244 MCP tools GitHub →

Gutierrez Public Lab

Social Modeling

Distribution Layer · Both Focuses

This site. Turns private experiments into public stories with real metrics. The distribution layer for everything else.

118 items, 282 R viz

Other Projects

Kiro CLI Factory Preferences

CLI Tooling · Learning Experiments

Factory for generating CLI tools with consistent patterns.

CLI generator framework GitHub →
Pirate Ship Preferences

Shipping Automation · Learning Experiments

Shipping label automation and rate optimization.

Automated shipping GitHub →

Pillar Mapping

Each project advances a pillar of the AI individuality thesis. Personality is the foundation; the others express it.

Project Domain Pillar Key Metric
Oil Model Commodity Futures Preferences +2.16pp Brier vs Polymarket
Jobs-Apply Career Automation Preferences 6 channels, 333+ applications
Dakka Agent Orchestration Persistence 7.8K lines Rust, Tauri v2
BloomNet Developer Intelligence Memory Analytics + context curation + alerts
RedCorsair Brand System Preferences Slate + Journal design tokens
Quick-Fin Financial Tooling Preferences 244 MCP tools
Gutierrez Public Lab Distribution Layer Social Modeling 118 items, 282 R viz