Architecture AI Updates: July 29, 2026

1. The Orchestrator’s Tax Reframes Subagents As Context Protection

Martin Fowler. Writing on martinfowler.com, Rahul Garg of Thoughtworks argues that the real value of subagents is protecting the orchestrator’s working memory rather than saving wall-clock time. He contends that every token in an orchestrator’s context competes for its attention, so isolating noisy intermediate reasoning inside worker agents keeps the main thread focused, and that tasks sharing the same mental model should stay together rather than be split across agents. He also warns that adding approval gates often masks missing information, favoring encoding single clarifying facts into standing rules over building bureaucratic decision procedures into orchestration workflows. Source