Field notes on the runtime reliability and cost of production LLM-agent systems — real traces, real invoices, real incidents. By Yevhen Herasimov.https://yevhenherasimov.com/enhttps://yevhenherasimov.com/blog/agentic-coding-harness-benchmark/https://yevhenherasimov.com/blog/agentic-coding-harness-benchmark/Same LLM, five coding harnesses, 120 runs: 80% of the token-cost gap between Claude Code, Codex CLI, aider, opencode and pi is scaffolding, not work.Tue, 04 Aug 2026 00:00:00 GMTcostevaluationtoolinghttps://yevhenherasimov.com/blog/two-claude-accounts-one-machine/https://yevhenherasimov.com/blog/two-claude-accounts-one-machine/CLAUDE_CONFIG_DIR moves the whole config directory, not just the login. Two directories, two shell functions, done.Sat, 25 Jul 2026 00:00:00 GMTtoolingcost