Me
RSS FeedExploring | Thinking | Connecting | Practicing
Featured
-
How I Collaborate with an Agent Team That Keeps Getting Better
From "agent teams aren't on this trajectory" to developing every day with an agent team that keeps getting better — how members become dimensional, why the member replaces the session as the default unit, where subagents should retreat to, and which judgments are, to this day, still only experience.
-
Focus: Regulating Exploration with 'Less is More'
Less is more. Don't let 'exploration' become an escape; exploration requires clear objectives and boundaries, not the suspension of judgment.
Recent Posts
-
Representative Projects from the Past Year
A review of several representative projects from the past year, and the judgments they helped me form about products, agents, context, and AI collaboration.
-
Notes on Agents, Tools, and Honesty
Collected notes from January to February 2026 — on message injection as a paradigm shift, single-agent parallelism over agent teams, AI coding and tool building, agent social networks, and how collaborating with AI demands honesty with yourself.
-
Context, Scaffolding, and Simplicity in Agent Design
Questioning the full-context approach for agents, exploring progressive disclosure vs. pre-built context, and reflecting on simplicity as a guiding principle in agent architecture and collaboration.
-
A Collection of Thoughts on Models, Context, and Agents
Reflections on recent AI model experiences, prompt engineering principles, context design strategies, and philosophical considerations about AI anthropomorphism and human-AI interaction.
-
Coding Model Usage and the Divergence and Convergence of Model Ecosystems
An exploration of coding model usage patterns, the differentiation and convergence of model ecosystems, and the evolving landscape of AI reasoning and agentic capabilities.