A sandbox limits where an agent can act when code, tools, or instructions go wrong. It is containment—not permission, judgment, or trust.
Open note
Technology, systems, publishing, and practical experiments.
A personal publication exploring developer workflows, agent systems, content infrastructure, tools, design, and the process of building useful things with care.
Featured deep dive
Why AI Feels Different When It Remembers You
Memory is part of the shift from one-off answers to ongoing continuity. That changes how AI systems work, how they feel, and what users need to control.
Read deep dive
Notes
View archiveGood automation gives your attention back. Bad automation trades saved clicks for more alerts, summaries, and systems to manage.
Open note
A beginner-friendly guide to the difference between what an AI learned during training and what it can retrieve from current sources.
Open note
Spotlights
View all
Buzz
Block’s fast-moving open-source workspace is becoming a credible home for agent teams, but its sharpest advantages over Discord still come with early-product costs.
Read spotlight
Hermes Agent
Hermes turns memory and experience into reusable skills. The promise is compelling, but durable learning also makes mistakes durable.
Read spotlightProjects
View all
Teaching Buzz to Mention Remote Agents
A self-hosted Buzz pilot exposed the work behind first-class remote agents: identity discovery, native Nostr mentions, isolated routing, and a clearer interoperability contract.
Read project
Expert System for Faculty Profiles
An early-stage expert-profile system built to improve evidence quality, reduce bluffing, and keep automated summaries aligned with real signals.
Read project
WordleBot
A PHP Discord bot that parses shared Wordle results with pattern matching, tracks performance over time, and turns daily puzzle posts into lightweight community stats and achievements.
Read project