Newsdesk
Engineering
Data & AI
Industries
Enterprise Systems
Go-to-Market
Longform
DesignIndiaAll stories

Craft & Leadership

14 November 2026

50 new posts across 8 practice and leadership sources — 7 worth your time.

One week of Craft & Leadership, 7 stories, as published.

Architecture

Presentation: Batch Intelligence at Scale: Cost-Efficient Multi-Agent LLM Workflows with Built-In Resilience

Walmart Global Tech deployed multi-agent LLM workflows across 4,700 stores for nightly inventory forecasting. The team resolved schema violations and network drops using MCP grounding while cutting cloud inference spend from over $100k to $10k per month by shifting from provisioned throughput to pay-as-you-go models.

Why it matters — Multi-agent batch workflows can reduce inference costs by 90% when moving from provisioned throughput to pay-as-you-go combined with tool-grounded error recovery.

InfoQ Architecture · 14 November 2026 · Read the original →

Testing & Quality

Tests as data: ensuring AI-generated code does what we want

Matteo Vaccari analyzes the shifting role of automated testing when AI agents generate both application code and unit tests. He argues that developers must shift focus from writing developer-level tests to crafting customer tests, which serve as the primary contract for agent-generated code.

Why it matters — When AI generates implementation and developer tests simultaneously, customer-level acceptance tests become the primary specification contract.

Martin Fowler · 14 November 2026 · Read the original →

Productivity

Your engineers got twice as fast. Your lead time only moved 4%

This article examines why doubling raw code generation speed with AI assistants yielded only a 4% movement in overall software delivery lead time. It argues that engineering organizations must focus on broader SDLC bottlenecks like code review and deployment rather than isolated coding velocity.

Why it matters — Optimizing code authoring speed yields diminishing returns if downstream processes like code review and release gates remain unoptimized.

LeadDev · 14 November 2026 · Read the original →

Practice

JioHotstar Benchmarks Knowledge Bases for AI Coding Agents

JioHotstar benchmarked six knowledge base retrieval techniques for AI coding agents across public benchmarks and an internal QA harness. The evaluation revealed that structural search excels at precise code tasks, while semantic retrieval is necessary for ambiguous requirements, highlighting key token cost and accuracy trade-offs.

Why it matters — Structural code search minimizes token usage for deterministic tasks, while semantic retrieval should be reserved for ambiguous, broad-context queries.

InfoQ Architecture · 14 November 2026 · Read the original →

Architecture

Long-horizon agents and durable execution

Kiran Prakash outlines the durable execution architectural pattern used to maintain long-running AI agent workflows that span days or weeks. The pattern enables agents to survive process crashes, infrastructure failures, and delayed external human inputs without losing execution state.

Why it matters — Durable execution preserves agent state across crashes and asynchronous external delays without manual state re-hydration logic.

Martin Fowler · 14 November 2026 · Read the original →

Leading Teams

Shipping faster didn’t save eBay – with Randy Shoup

Randy Shoup draws on four decades of engineering leadership experience at eBay, Google App Engine, and Stitch Fix to discuss engineering velocity and architecture. He addresses why accelerating code delivery fails to rescue businesses without clear technical strategy and architectural alignment.

Why it matters — Increased delivery speed amplifies underlying strategic and architectural flaws if organizational direction remains misaligned.

Refactoring · 14 November 2026 · Read the original →

Leading Teams

One-way and two-way doors

This post examines how to maintain effective decision-making frameworks alongside modern open-communication management cultures. It addresses the distinction between reversible and irreversible decisions to establish lightweight governance that doesn't bottleneck teams.

Why it matters — Categorizing decisions by reversibility prevents low-risk choices from getting stalled in management review processes.

The Engineering Manager · 14 November 2026 · Read the original →

At least the AI can wait for approval as patiently as the rest of us.

7 stories, every Saturday

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

← Previous issue Next issue →