0x0011f60 · SEC/RESOURCES · REV.03 · 12 articles

Notes from the bench.

September 30, 2026
11 min

The Computer Was Always Programmable. AI Just Removed the Excuse.

The GUI is a menu someone else wrote. The CLI, cloud, APIs and AI expose what a computer can actually be made to do, and knowing that map is the skill.

Computational Agency→

September 3, 2026
8 min

Shipping to TestFlight Without Opening Xcode

A Python terminal game shipped to TestFlight as a native iOS app with the Xcode GUI never opened once: Windows editing, a Mac mini over SSH, 175 tests green.

Swift→

August 23, 2026
9 min

The Agent Disobeyed and Was Right

An agent extended an existing renderer instead of writing a throwaway. Five uncoordinated inputs explain the happy accident and suggest testable principles.

AI Architecture→

July 12, 2026
6 min

Confidence Is Not Evidence: 91 Fabrications at Certainty 1.00

My self-hosted memory system's LLM classifier filed 91 reconciliation proposals last week. Every one was fabricated, some at confidence 1.00. Containment held: zero bad writes.

AI Architecture→

July 12, 2026
6 min

144.8 Seconds to 11.8

A Rust disk analyzer felt slow, so a multi-agent review swept the whole stack. The walk that looked parallel wasn't, every file cost three kernel round-trips, and the NTFS fast path was throttled by a wrapper crate reading 4 KB per syscall. Same volume, same totals: 144.8 seconds down to 11.8.

Rust→

July 6, 2026
6 min

93.9% Savings, Compared to What?

My context optimizer reported 93.9% token savings. Claude Code flagged the same tool as my biggest context consumer. Both were true. Here's what honest measurement took.

AI Architecture→

July 6, 2026
7 min

The Memory Outlives the Model

A thought experiment (an AI companion for my own kids) led to a conclusion about all personal AI: memory is the persistent layer, and the keys belong to whoever it's about.

AI Architecture→

July 3, 2026
6 min

Coordination Debt: The 116 KB File Every Agent Had to Read

An audit of a multi-agent ticket-to-ship pipeline found the real cost wasn't model reasoning. Three coordination files had bloated into history logs: 116 KB down to 5 KB.

Multi-Agent Systems→

February 10, 2026
4 min

Orchestration Decides Who Thinks

Part 3 of 3 on AI systems that scale. Uncontrolled cognition is the real cost center: orchestration decides who thinks, and memory turns cognition into a compounding asset.

AI Architecture→

February 9, 2026
4 min

Swarm Architecture: Why Flat Authority Fails

Part 2 of 3 on AI systems that scale. Most multi-agent systems fail from flat authority and no gating: disciplined swarm design turns agents into compounding intelligence.

AI Architecture→

February 8, 2026
4 min

Intelligence Is Not the Bottleneck

Part 1 of 3 on AI systems that scale. Most failures trace to missing architecture, not weak models: leverage comes from controlling cognition, not upgrading tiers.

AI Architecture→

August 2, 2025
3 min

The System Prompt is the Strategy

A system prompt is not branding copy. It is the layer that decides how an agent resolves ambiguity, refuses, and falls back. Write it like firmware.

Prompt Engineering→

Written by the operator, from production work.

12 of 12