Making Your Web Page AI Friendly
Seven ways to make website content and functionality accessible to AI agents, covering JSON-LD, Markdown, JSON catalogs, XML sitemaps, OpenAPI, MCP tools, and WebMCP.
Read on Medium →Insights
Writing on architecture, AI execution systems, and dependable delivery.
New Insights
nSpace Insights
The Terminology Is New. The Architecture Isn't.
Prompt engineering, loop engineering, graph engineering — and why governance was always the missing node.
Read insight →Why Abracapocus Beats /goals for Real Agentic Coding
Why bounded tasks, fresh context, evidence-backed handoff, and cost-aware routing make Abracapocus better suited than single-goal loops for real agentic coding.
Read insight →Governed Async Execution in Abracapocus
How Abracapocus runs concurrent AI coding work under enforced safety constraints with durable evidence attached to every task.
Read insight →Tiered Retry Is a Failure Economy
Why blind retries waste model spend, and how diagnosis, fresh context, targeted repair, and evidence-based escalation make agentic coding cheaper and more reliable.
Read insight →The Economics of Abracapocus
Why governed execution will matter as AI execution costs become explicit and inference stops being treated as free.
Read insight →Abracapocus Building Abracapocus
Abracapocus is being developed through the same governed autonomous execution model it proposes for production software systems.
Read insight →Beyond the Dark Software Factory
Dan Shapiro’s Five Levels describe the path toward the dark software factory. Abracapocus points toward the governed software factory beyond it.
Read insight →From the Karpathy Loop to Governed Autonomous Development
AI coding is moving from conversational iteration to governed execution through contracts, verification, evidence, and bounded autonomous delivery systems.
Read insight →Dependable AI-Assisted Development Series
External article series from foxmike.github.io.
AI-Assisted Development Scales Through Verification, Not Generation
Read external article →
The Unreliable Narrator Problem in AI-Assisted Development
Read external article →
Stage Separation: Why Software Construction Needs a Supervisor Pattern
Read external article →
Task Contracts as the Unit of Work in AI-Assisted Development
Read external article →
Multi-Backend Routing Is Insurance Against Model Risk
Read external article →
Tiered Retry Is a Failure Economy
Read external article →
Minimal Blast Radius Is What Makes AI Automation Tolerable
Read external article →
Curated Task Execution Beats Dynamic Planning in Real Work
Read external article →
Anti-Patterns in AI-Assisted Development
Read external article →
State, Resumability, and Why Chat Memory Is the Wrong Primitive
Read external article →
External Notes
Older external notes on applied AI and data workflows.
CNN for Image Classification
Why it matters: demonstrates how robust image pipelines can classify operational inputs consistently and support faster downstream decisions.
Read on Medium →Extracting Structured Data from Images with GPT-4 Vision + Instructor
Why it matters: shows a practical method for converting visual documents into structured, machine-usable data with controlled output formats.
Read on Medium →Sharing Data with Large Language Models
Why it matters: explains how to share business data with language models while preserving context, controls, and decision quality.
Read on Medium →