Now

What I'm focused on this quarter. Updated periodically.

Last updated: April 2026

Writing

• Expanding the MCP section with more practical server examples
• Writing up my agent evaluation methodology
• Drafting a deep dive on safety boundaries for autonomous agents

Research

• Watching how the MCP ecosystem matures, which servers actually get adopted vs, which stay on the shelf
• Stress-testing the limits of Claude's extended thinking on multi-hour autonomous tasks
• Studying failure modes of multi-agent orchestration (spoiler: there are many)

Reading

• Every new Anthropic release post
• The latest ReAct and CoT ablation papers
• MCP spec updates and RFCs

What I'm not doing

• Not chasing the next fine-tuning benchmark
• Not building a training pipeline
• Not starting an agent-as-a-service SaaS (the space is saturated)
• Not writing yet-another "why AGI is near" essay

Inspired by the /now page movement.