AI News from Towards AI

Latest coverage from Towards AI, summarized and scored for signal.

• From Prompt Engineering to Harness Engineering: The Next Evolution of LLM Systems — Gradient-based attribution in transformers systematically mislabels component importance: early-layer "Gradient Bloats" dominate rankings despite negligible function while late-layer "Hidden Heroes" are undervalued — rank correlation collapses to ρ = -0.18 in some seeds, challenging a core assumption of mechanistic interpretability.