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 →Insights
Writing on architecture, AI execution systems, and dependable delivery.
nSpace Insights
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 →