SlopCodeBench: coding agents still break old code
Examine October 2026 SlopCodeBench results: Astra and Opus scores, XHigh costs, regressions, and practical changes for agent-assisted development.
BLOG
Explore technical articles and guides about testing.
Examine October 2026 SlopCodeBench results: Astra and Opus scores, XHigh costs, regressions, and practical changes for agent-assisted development.
Explore Pairing Matters and its 1,800 coding-agent runs: compare models and harnesses, understand failure modes, and measure useful token efficiency.
A comprehensive technical deep dive into the Superpowers methodology, exploring its subagent orchestration, strict TDD enforcement, and how it transforms AI from a chaotic coder into a disciplined engineering partner.