https://genalphai.com/ Research-driven AI engineering. Depth over hype.enhttps://genalphai.com/context-rot-and-the-dumb-zone/ https://genalphai.com/context-rot-and-the-dumb-zone/Wed, 10 Jun 2026 23:17:06 GMT Context rot degrades LLM agents well inside advertised windows. Why the ~100k dumb zone exists, what 'lost in the middle' research shows, and the inner-loop/outer-loop architecture that fixes it.https://genalphai.com/swe-bench-pro-vs-verified/ https://genalphai.com/swe-bench-pro-vs-verified/Wed, 10 Jun 2026 22:46:07 GMT OpenAI deprecated SWE-bench Verified after finding flawed tests in 59.4% of hard tasks. How SWE-bench Pro and DeepSWE's 32.5% verifier error rate change agent evaluation.https://genalphai.com/agents-md-vs-claude-md/ https://genalphai.com/agents-md-vs-claude-md/Wed, 10 Jun 2026 22:04:13 GMT AGENTS.md, CLAUDE.md, and .cursor/rules compared: three-tier permissions, context budgeting, and the canonical-plus-adapters pattern that keeps coding agents obedient.https://genalphai.com/ralph-wiggum-loop-stateless-agents/ https://genalphai.com/ralph-wiggum-loop-stateless-agents/Wed, 10 Jun 2026 21:42:26 GMT The Ralph Wiggum loop re-feeds one prompt to a fresh agent process forever, using files and git as the only memory. Why this dumb pattern keeps winning.https://genalphai.com/reasoning-first-llms/ https://genalphai.com/reasoning-first-llms/Wed, 10 Jun 2026 20:50:22 GMT LLMs rationalize answers they already chose. Process supervision, self-consistency, and faithfulness probes force models to reason to the right answer.