08 August 2026

48 new posts across 8 practice and leadership sources — 5 worth your time.

One week of Craft & Leadership, 5 stories, as published.

Presentation: Rewriting All of Spotify's Code Base, All the Time

Spotify developed an AI coding agent named 'Honk' to manage complex, fleet-wide codebase migrations across thousands of repositories. The post outlines key architectural decisions, such as decoupling CI verification runtimes from AI agents and managing automated pull request bottlenecks.

Why it matters — Decoupling CI verification runtimes from AI agents prevents automated pull request bottlenecks during large-scale codebase migrations.

InfoQ Architecture · 08 August 2026 · Read the original →

Incidents & Ops

Wiz Discloses CosmosEscape, and Practitioners Debate What Customers Could Have Done

Wiz Research identified a vulnerability chain called CosmosEscape that allowed escaping Azure Cosmos DB's Gremlin sandbox to access a platform-wide read/write key. While Microsoft patched the entry point in two days, it took until July 2026 to completely remove the compromised key, sparking industry debate over shared responsibility and rearchitecture costs.

Why it matters — Sandbox escapes can expose platform-wide master keys, demonstrating that cloud provider mitigation timelines can extend to several years and require significant customer-side rearchitecture.

InfoQ Architecture · 08 August 2026 · Read the original →

Architecture

The Orchestrator's Tax

This post discusses how subagents should be utilized to protect an orchestrator's working memory and context window. Rather than focusing solely on parallel execution or time saved, subagents serve to offload reasoning tasks so they do not compete for the orchestrator's attention.

Why it matters — Subagents should be treated as tools to protect the orchestrator's context window and working memory by offloading unnecessary reasoning.

Martin Fowler · 08 August 2026 · Read the original →

Practice

The Economic Benefit of Refactoring

This post details an experiment by Giles Edwards-Alexander investigating whether decomposing a large function can reduce LLM token costs. The findings suggest that refactoring code structure can now yield direct, measurable economic benefits.

Why it matters — Decomposing large functions can directly reduce LLM token costs, providing a measurable financial incentive for code refactoring.

Martin Fowler · 08 August 2026 · Read the original →

Careers

Presentation: Getting Rid of LeetCode Interviews in the World of AI

Daniel Doubrovkine discusses why traditional whiteboard LeetCode interviews fail to accurately evaluate senior engineering talent. He proposes alternative interview frameworks that focus on evaluating human judgment, system design, and hands-on AI collaboration.

Why it matters — Shifting interview loops from algorithmic whiteboard tests to evaluating human judgment and hands-on AI collaboration provides a better hiring signal for senior talent.

InfoQ Culture · 08 August 2026 · Read the original →

At least we can now use AI to generate the code we will rewrite next sprint.

5 stories, every Saturday

Published here every week. Follow by RSS to get it as it lands.

Subscribe by RSS All issues

Next issue →