Blog
Field notes on AI, engineering leadership, and building things that work.
Anthropic Mythos: What 3,000 Leaked Documents Actually Tell Us
Two separate Anthropic leaks in five days. A new model tier above Opus. An always-on agent that dreams at night. What Mythos, KAIROS, and 512K lines of source code reveal about the next generation of AI.
nb_researchnb_anthropicnb_agents
I Built a Database MCP Server Because Existing Ones Drove Me Crazy
The behind-the-scenes story of building a2db: what broke with dbhub, the bugs I hit along the way, and what I learned about MCP server development.
iorlas_toolsiorlas_agentsiorlas_mcp
How AI Agents Actually Remember Things
A look at persistent memory architectures for AI agents — what works, what doesn't, and why most approaches fail in production.
aiagentsarchitecture
The Open Memory Problem: Why AI Agents Still Can't Remember Properly
Eight agent memory systems compared. Everyone solves storage. Nobody solves intelligence.
iorlas_researchiorlas_agents