All Blogs

When AI systems start building long-term memory, local-first stops being just a privacy choice — it becomes a question of ownership, trust, and continuity.

A reflection on trains, mountains, family, and the quiet reset that comes from slowing down.

A reflection on effort, endurance, and why some difficult paths stay with us.

AI remembers context, but it still doesn’t learn people.

Agents don’t get better just by remembering more — they need a loop that changes what they do next.

Most agent demos work because everything goes right — production is where systems are tested on everything that doesn’t.

Building memory for AI agents turns out to be less about storage and more about interpreting what users really meant.

Agentic AI isn’t just about bigger models. It’s about systems that survive real-world complexity.