Building SoPhi: a Kotlin Agent Harness

The engineering behind SoPhi, one decision at a time. Also published on Medium and Towards AI.

1. #1 Why We Built a Kotlin Agent Harness Instead of Using One20 Jul 2026
2. #2 Designing SoPhi: Eight Modules and the Decisions Behind Them22 Jul 2026
3. #3 The Agent Loop: a First-Principles Design24 Jul 2026
4. #4 Provider Abstraction Without Framework Lock-in30 Jul 2026
5. #5 The memory that can’t lie: crash-safe sessions with nothing but JSONL11 Aug 2026
6. #6 Hands that Don’t Freeze: A Tool Interface Built For a Future it Hadn’t Earned Yet30 Sep 2026
7. #7 From Inert Files to Self-Writing Skills: Evolving an AI Agent’s Skill System7 Oct 2026
8. #8 Calling in a subcontractor: subagents as just another Tool19 Oct 2026