Blog

Evidence-based retrospectives for human-agent collaboration. A feedback loop where both humans and agents learn from each session.

When Docker, Apple, and every hyperscaler builds microVM sandboxes, containers aren't enough. A deep dive into the AI sandbox landscape, the isolation technologies behind them, and why the walls need to be real.

Building better experiences for developers and AI agents working together. Our vision for the future of collaborative software development.

A pattern for managing 20+ git repositories across work projects, side projects, infrastructure, and experiments using a central planning repository with unified documentation, cross-project orchestration, and shared AI agent configuration.

GLM-4.7-Flash closes the gap between local and cloud AI coding tools, achieving 59.2% on SWE-bench while running on consumer hardware under MIT license.

Claude Cowork is an autonomous desktop agent from Anthropic that extends Claude Code capabilities to general knowledge workers through a visual interface in the Claude Desktop app.

A Claude Code plugin for consuming and curating YouTube videos with AI-generated summaries, key takeaways, and structured transcripts.

Combining spec-driven development with backlog.md for a structured AI-augmented development workflow that keeps task status and work produced together in the git repository.