The Bike4Mind Blog
Field notes on owning your AI — models, agents, and the open stack.
★ Featured
The Mu Strategy: How to Build on Hyperscalers Without Being Owned By ThemNovember 17, 2025The Norway essay showed hyperscalers running a sovereign-grade macro trade. This essay explains the third option: Mu — how to use hyperscalers as infrastructure (not gods) while keeping your strategic freedom. Rent their muscles. Own your brain. Build where they structurally cannot follow.AI
Architecture
Cloud & Infrastructure
Product Strategy
Bring the Open Back to AIJuly 4, 2026On July 4th we opened Bike4Mind’s core: source-available under BSL 1.1, converting to Apache-2.0 on a two-year clock written into the license. Why the closing frontier made it necessary — and why the car keeps driving when the engine gets pulled.
AI
Open Source
Business
Bike4Mind
The License MazeJuly 3, 2026Walking every exit of the open-core licensing maze — MIT, AGPL, SSPL, BSL — and why the door out is a generous grant with a two-year Apache ratchet you cannot stop.
AI
Business
Strategy
Open Source
Beauty Is the RewardJune 28, 2026Why a flower, a proof, and a face all feel the same — beauty as the felt signal of a successful compression — and why that feeling can lie. The capstone of a four-part series on intelligence.
AI
beauty
aesthetics
intelligence
Distillation Attacks on the UniverseJune 28, 2026Why a 20-watt brain out-learns a megawatt model — and what it stole from a billion-year pretraining run. The capstone of a trilogy on eyeballs, leylines, and cheap intelligence.
AI
AGI
intelligence
neuroscience
Give the Model EyeballsJune 28, 2026If an agent can clearly see, measure, or hear a task, it's 80% solved — the mechanism, where it betrays you, and why it's the first thing I screen for when I hire.
AI
agents
LLM
AGI
LeylinesJune 28, 2026Imagination is the bottleneck now. My hunch: it's a search across a high-dimensional space along ridges that were already there — and measuring them builds the last eyeball.
AI
imagination
AGI
philosophy
How We Built /think-tank — A Skill for Infographics That Can't LieJune 18, 2026The story of building a Claude Code skill that turns data into honest, proportional infographics — born from one stubborn water drop that refused to lie.
AI
Claude Code
Data Visualization
Tools
A Skill Is a Voice in Your Agent’s Ear: How to Safely Vet One Before You Run ItJune 14, 2026How to git clone and review an AI skill, and why you must sever its phone-home channel even when it looks completely benign.
AI
Security
AI Alignment
Leadership
Two Days, Two Codebases, Fourteen FindingsApril 10, 2026How an agentic security tool I built three days ago adapted to a brand-new target in under an hour, found 14 issues, and produced a clean fix branch in three hours total.
AI
Security
Claude Code
Agentic
A Thermodynamic Learning Algorithm from 1995 Still Works to learn Kanji Fast!March 16, 2026From a 1995 algorithm paper to a live product at k2kanji.com — 12,000 lines, 9 deploys, and my wife and I testing it together. This is what strong AI actually delivers.
ai
kanji
japanese
learning
One Image, One Night: Building Strait SweeperMarch 15, 2026How a single meme image became a fully playable Minesweeper game set in the Strait of Hormuz — built and shipped in one evening with Claude Code.
Game Design
AI
Shipping
Building a Prototype-to-Spec Pipeline with Claude Code and PlaywrightMarch 12, 2026How we built an AI skill that turns a CPO's HTML prototype into an 820-line implementation spec in 20 minutes — and why reading the source matters more than clicking the buttons.
AI
Engineering
Product Development
Claude Code