EvalTrim 0.6.0 turns AI-agent eval maintenance into an evidence-driven workflow. Instead of only running evals, it identifies redundant tests, finds unique behavioral witnesses, simulates removal before recommending maintenance, tracks regressions and evaluation debt, and protects critical coverage. It runs locally and never silently deletes tests.
Malaysia Linguistics Lab is an interactive project exploring four Malaysian minority and Indigenous languages: Bidayuh, Mah Meri, Kadazan-Dusun, and Iban. It combines language learning, a World Explorer, dictionary, quizzes, learning paths, heritage, language comparison, achievements, and an AI Tutor. What began as a simple website gradually grew through continuous testing, design improvements, and trial and error.
