News Around the World — 2026-10-28

AI

• What Effort Level Should I Choose? On the Economics of Reasoning Effort in LLM Agents — Across four benchmarks and both proprietary and open-weight models, higher reasoning effort often fails tasks lower effort solves and costs up to 42x more on easy tasks; escalating effort only after a verified failure matches or beats top-effort accuracy at lower cost for most proprietary models.
• HABIT-Bench: Evaluating Habit Induction from Longitudinal Weak Evidence in Agent Memory — Testing eight LLM agent memory methods on inferring user habits from weak evidence across hundreds of sessions in four domains, the best method reaches only 33.61% macro accuracy on four-choice tasks, even when annotated evidence is supplied directly.
• The October inflection point for vibe game development — AI-driven “vibe coding” of computer games reached a turning point in October 2026, mirroring the inflection point coding agents hit in November 2025.