New here?
Read this first →
What this is, how it got here, and why I'm not apologizing for any of it.
Build the thing that builds the things.
negrini.io is a public notebook on Agentic Harnesses: the infrastructure around a base model that turns ad-hoc prompting into repeatable, auditable work.
You’ll find threads on context and memory, orchestration, knowledge graphs, and a technical memoir exploring the ADHD–AI parallel.
This is a living system. Artifacts are shipped in public—some are draft, some are revising, some are solid.
Choose an entry point
ADHD–AI Parallel
Context windows are working memory. Reframing ADHD as architecture—and building external systems that make both humans and agents work.
Infrastructure First
Stop improvising. Build the factory: context, workflows, and quality gates that create compounding returns.
Agentic Context Engineering
Context is not a chat transcript. It’s a compiled view over state: memory tiers, summarization, and retrieval that keep agents coherent.
Knowledge Graph Architecture
External memory that works by meaning, not keywords—semantic search, GraphRAG, and relationship-aware retrieval.
How this site is organized
Artifacts
Long-form notes, specs, and build logs.
Threads
Themes that connect ideas across the knowledge base.
Book
An in-progress technical memoir—follow a reading path.
Now
Current focus areas and active builds.
Trail
A guided journey from marketing to AI architecture.
Glossary
Plain-language and technical definitions.
Recommended starts
Browse all artifactsAgentic Harness Explainer Script
Quick explainer: 3-4 minute narration script for Claude Code Harness...
Claude Code Harness Builder
Specification for building Claude Code plugin harness that wraps GSD workflow with...
ADHD Brain Parallels with AI Agents
Core thesis document mapping cognitive patterns to agent architectures
GraphRAG System Capabilities Overview
Comprehensive overview of graph-based RAG capabilities