Journal May 19, 2026
Gutierrez Public Lab
Building the infrastructure for AI individuality
AGI capabilities exist. Individual AI behavior does not. I'm systematically solving the building blocks: persistence, memory, preferences, and personality. Every experiment documented. Every failure published. Every win verified with real metrics.
Brier vs Polymarket
+2.16pp
Oil model
Data Quality
0.98 DQI
Autonomous ratchet
Tests Passing
607
Across 3 Rust repos
Active Channels
6
Job automation
Learning Experiments
Hypothesis-driven work across 64 experiments. 43 confirmed, tracked with research-lab rigor.
Browse Experiments →Self-Improving Agents
Systems that make themselves better: self-healing tests, self-managing context, self-improving data models.
See the Stack →Featured Projects
Featured Project 24 metrics, 2,740 sessions
Agent MQI
Composite quality score for AI coding assistants. 56.4% of my words were negative. The more I cursed, the better the code got.
Featured Project +2.16pp Brier vs Polymarket
Oil Model
Monte Carlo pricing engine for WTI crude. Seven ratcheted parameters. Beats the prediction market.
Featured Project 6 channels, 333+ apps
Jobs-Apply
Automated job applications across LinkedIn, Greenhouse, Lever, Workday, Direct, and Indeed with CDP browser control.
Featured Project 7.8K lines Rust
Dakka
Parallel Claude Code orchestrator with live API rate limit monitoring and spawn/kill UI. Ported from TypeScript to Rust.
Experiments
64
43 confirmed
Weekly Reviews
19
story-driven articles
Active Projects
5
public codebases
Journal Days
132
days documented
Latest
Experiment confirmed bloomnet May 4, 2026
Guardrail architectures combining hard-block hooks with memory rule...
Experiment inconclusive jobs-apply Apr 27, 2026
ToFu visitors on About are ready for BoFu action without more mid-f...
Experiment inconclusive jobs-apply Apr 27, 2026
Changelog readers who see shipping momentum respond to urgency framing
Experiment inconclusive jobs-apply Apr 27, 2026
Contact page visitors respond better to direct download than gentle...
Experiment inconclusive jobs-apply Apr 27, 2026
Concrete metrics build more credibility than broad feature claims
Experiment inconclusive jobs-apply Apr 27, 2026
More frequent CTAs mid-scroll catch visitors before bounce, shorter...
Experiment inconclusive jobs-apply Apr 27, 2026