# Elevate Addy Osmani's newsletter on elevating your effectiveness. Join his community of 600,000 readers across social media. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## Brownfield Agentic Engineering DevFeed: [Brownfield Agentic Engineering]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-09-14T14:32:06Z Content type: article Language: en Sources: [Elevate]() Topics: [coding](), [Testing](), [systems](), [Artificial Intelligence](), [Software]() Tags: [agentic](), [agents](), [code](), [coding](), [engineering](), [software](), [systems](), [technical](), [testing]() ### AI overview The article discusses agentic engineering in older brownfield codebases, where important constraints and system behavior exist outside the repository. It emphasizes making hidden constraints visible, using strong testing and verification, and introducing autonomous changes carefully to avoid brittle tests and technical debt. ### Source excerpt What it takes to run agents in a codebase older than the team ## Agentic Skill Decay DevFeed: [Agentic Skill Decay]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-08-31T14:30:37Z Content type: opinion Language: en Sources: [Elevate]() Topics: [Learning](), [Software Engineering](), [Development]() Tags: [agentic](), [agents](), [engineering](), [learning](), [software-engineering]() ### AI overview The article argues that coding agents can complete tasks while bypassing the deliberate practice that builds expertise and judgment. It recommends forming hypotheses, understanding domains, reading diffs, debugging, reviewing code, and deliberately practicing decision-making, specification, steering, and verification. ### Source excerpt Agents can finish the task without teaching you anything. Building expertise now has to be deliberate. ## Auditing Coding-Agent Configuration, Skills, Memory, and Hooks DevFeed: [Auditing Coding-Agent Configuration, Skills, Memory, and Hooks]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-08-27T14:31:10Z Content type: tutorial Language: en Sources: [Elevate]() Topics: [Agent Skill](), [configuration](), [AI-assisted coding](), [Claude](), [sdlc]() Tags: [agent](), [agent-skill](), [agent-skills](), [claude](), [coding-agents](), [configuration](), [markdown](), [sdlc]() ### AI overview This practical guide argues that coding-agent configuration files can become outdated as models, harnesses, and codebases change. It recommends periodically reviewing configuration, memory, skills, and hooks, while acknowledging concerns about token costs, excessive instructions, and uncertain effects on quality. The author describes using Claude's /doctor and maintaining Agent Skill packages. ### Source excerpt A practical guide to auditing what your coding agent still needs. ## Human judgment doesn't leave the software factory. It relocates. DevFeed: [Human judgment doesn't leave the software factory. It relocates.]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-08-21T14:31:12Z Content type: article Language: en Sources: [Elevate]() Topics: [Software](), [Testing](), [event driven](), [Architecture & Design](), [GitHub](), [pull-requests](), [mutation-testing](), [Security]() Tags: [agent](), [architecture](), [claude-code](), [codex](), [event-driven](), [github](), [mutation-testing](), [pull-requests](), [quality](), [security](), [software](), [testing]() ### AI overview This field guide explains how to build a repeatable software factory while preserving human judgment and ownership. It recommends human involvement in product intent, system design, quality standards, code review, and merge decisions, supported by automated checks and deliberate experimentation. The article argues that a factory is most useful when work must be repeatable and event-driven across queues such as GitHub issues, Slack, Linear, or a backlog. ### Source excerpt A field guide to building a software factory that still has an owner. ## Practical Loop Engineering DevFeed: [Practical Loop Engineering]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-08-14T14:30:53Z Content type: opinion Language: en Sources: [Elevate]() Topics: [Loop Engineering](), [AI Agent](), [Claude Code](), [codex](), [Code review](), [Scripting, bash]() Tags: [agent](), [agents](), [bash](), [claude-code](), [code-review](), [codex](), [engineering](), [loops](), [workflows]() ### AI overview An opinion article about practical loop engineering for AI-assisted software development. It describes running multiple agents in parallel, delegating bounded tasks with explicit stopping conditions, and using goal- and timer-based loops in Claude Code and Codex. It emphasizes careful supervision and well-defined goals and constraints, especially for complex or high-risk codebases. ### Source excerpt Goals, loops, and the discipline of not delegating your judgment ## Agentic Code Quality DevFeed: [Agentic Code Quality]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-08-08T14:31:15Z Content type: opinion Language: en Sources: [Elevate]() Topics: [Code quality](), [Testing](), [mutation-testing](), [Code review]() Tags: [agentic](), [code-quality](), [code-review](), [mutation-testing](), [testing]() ### AI overview The article argues that code quality for agent-generated software depends on constraints enforced by the surrounding harness, environment, and operating system. It describes quality gates including unit, property, and acceptance tests, mutation testing, code-quality metrics, and checks on proposed changes, while noting risks from autonomy, ambiguity, brittle environments, nondeterministic builds, missing permissions, and weak tests. ### Source excerpt Quality now depends on the constraints you set around your agents. ## Software Factories, Light and Dark DevFeed: [Software Factories, Light and Dark]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-07-22T03:44:07Z Content type: opinion Language: en Sources: [Elevate]() Topics: [Loop Engineering](), [Software](), [Artificial Intelligence]() Tags: [agentic](), [agents](), [harness](), [loops](), [software]() ### AI overview The article revisits the idea of software factories through loops, harnesses, and factories. It distinguishes light factories, where humans remain involved, from dark factories, where agents scope, build, and ship code with limited human review. It argues that autonomy requires carefully designed checks, boundaries, tools, memory, and completion gates. ### Source excerpt A software factory is harnessing loops at scale. ## Why loop engineering needs a human at the boundary DevFeed: [Why loop engineering needs a human at the boundary]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-07-09T14:31:24Z Content type: opinion Language: en Sources: [Elevate]() Topics: [Loop Engineering](), [Software](), [systems]() Tags: [agentic-engineering](), [agents](), [engineering](), [observability](), [process](), [software](), [systems](), [trust](), [verification]() ### AI overview The article argues that engineers must own the outer loop of agentic software systems. It defines quality checks, production verdicts, and answerability as human responsibilities, while describing agents as models combined with tools, memory, permissions, sandboxes, tests, observability, and recovery. ### Source excerpt Why loop engineering needs a human at the boundary ## Agentic Autonomy Levels DevFeed: [Agentic Autonomy Levels]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-07-03T14:30:33Z Content type: article Language: en Sources: [Elevate]() Topics: [Artificial Intelligence](), [Orchestration](), [Agent Skill](), [Claude](), [Claude Code](), [codex]() Tags: [agent](), [agentic](), [agentic-engineering](), [agents](), [ai](), [claude](), [claude-code](), [codex](), [delegation](), [multi-agent](), [orchestration](), [parallel](), [scopes](), [trust](), [verification]() ### AI overview This article proposes a two-axis model for evaluating agentic engineering: agency, describing how independently a single agent operates, and orchestration, describing the skill of coordinating multiple agents. It argues that a single autonomy ladder is insufficient for multi-agent work and emphasizes matching autonomy levels with appropriate verification. ### Source excerpt A working model of autonomy for agentic engineering ## Agentic Code Review DevFeed: [Agentic Code Review]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-06-16T14:31:22Z Content type: article Language: en Sources: [Elevate]() Topics: [Code review](), [AI-assisted coding](), [Pull Request](), [Claude Code](), [codex]() Tags: [agentic](), [agentic-code-review](), [claude-code](), [code-review](), [codex](), [coding-agents]() ### AI overview This article argues that coding agents have shifted the main engineering constraint from writing code to deciding whether generated changes are trustworthy. It describes using Claude Code and Codex to triage incoming pull requests while emphasizing that review needs vary by developer and team context. ### Source excerpt Why review is the most leveraged skill in software right now. ## Loop Engineering DevFeed: [Loop Engineering]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-06-08T14:31:21Z Content type: article Language: en Sources: [Elevate]() Topics: [Loop Engineering](), [AI-assisted coding](), [Claude](), [Claude Code](), [Agent Harness](), [codex]() Tags: [agent](), [agents](), [ai](), [claude-code](), [codex](), [coding-agents](), [engineering](), [harness](), [loops]() ### AI overview The article introduces loop engineering: designing systems that repeatedly prompt, check, record, and direct coding agents instead of manually prompting them turn by turn. It describes the approach as early and potentially costly in token usage, and compares its building blocks with features in Codex and Claude Code. ### Source excerpt Loop engineering is replacing yourself as the person who prompts the agent. ## Long-running Agents DevFeed: [Long-running Agents]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-04-30T14:30:29Z Content type: opinion Language: en Sources: [Elevate]() Topics: [AI Agent](), [Persistence](), [context window](), [Frontier Model](), [Process]() Tags: [agents](), [ai-agent](), [context-window](), [frontier-model](), [persistence](), [process]() ### AI overview This article explains long-running AI agents: systems that continue making progress across multiple sessions, context windows, and sandboxes while preserving state, recovering from failures, and leaving artifacts for later sessions. It distinguishes long-horizon reasoning from long-running execution and discusses the engineering challenges involved. ### Source excerpt A long-running AI agent can keep making progress over hours, days, or weeks. ## Architectural Bets for More Governable and Reliable Production Agent Systems DevFeed: [Architectural Bets for More Governable and Reliable Production Agent Systems]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-04-18T17:16:43Z Content type: opinion Language: en Sources: [Elevate]() Topics: [Architecture & Design](), [Security](), [Authorization](), [systems]() Tags: [agent](), [agentic](), [api](), [audit](), [autonomous](), [governance](), [identity](), [incident](), [pii](), [security]() ### AI overview This opinion article argues that production agent systems face a stack ceiling caused by fragile session logic, shared credentials, weak governance, and unreliable downstream dependencies. It proposes architectural changes, beginning with giving agents distinct platform-level identities instead of shared service accounts or inherited OAuth tokens. ### Source excerpt The bet every serious developer needs to make on on their agent stack ## Agent orchestration is replacing the editor as the center of developer work DevFeed: [Agent orchestration is replacing the editor as the center of developer work]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-03-20T14:31:23Z Content type: opinion Language: en Sources: [Elevate]() Topics: [Orchestration](), [ide](), [coding](), [Command-line interface]() Tags: [agent-orchestration](), [agents](), [claude](), [codex](), [cursor](), [developer](), [github-copilot](), [ide](), [terminal](), [workflows]() ### AI overview The article argues that developer work is shifting from line-by-line editing in an IDE toward supervising agents that plan tasks, modify files, run tests, and propose changes for review. It describes a workflow of specifying intent, delegating, observing, reviewing diffs, and merging, while noting that IDEs may evolve rather than disappear. ### Source excerpt How Agent orchestration is replacing the editor as the center of developer work ## 14 More lessons from 14 years at Google DevFeed: [14 More lessons from 14 years at Google]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-02-12T15:30:32Z Content type: article Language: en Sources: [Elevate]() Topics: [Google](), [systems](), [Code]() Tags: [advice](), [engineering](), [google](), [systems](), [teams](), [trust]() ### AI overview An article sharing lessons from 14 years at Google, focused on teams, decision-making, coordination, and the systems around engineering work. It emphasizes choosing the right problems and defining the decision needed from a meeting. ### Source excerpt This time about teams, trust, and the systems around the code. ## The 80% Problem in Agentic Coding DevFeed: [The 80% Problem in Agentic Coding]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-01-28T17:20:53Z Content type: opinion Language: en Sources: [Elevate]() Topics: [ai-coding](), [Artificial Intelligence](), [coding](), [Claude Code]() Tags: [agentic](), [agentic-coding](), [ai](), [ai-coding](), [architecture](), [bugs](), [claude-code](), [coding](), [developers](), [models]() ### AI overview The article argues that AI-assisted coding has shifted from completing roughly 70% of a task to handling 80% or more for some new or personal projects. It says the main difficulties have changed from syntax errors to conceptual failures, including unchecked assumptions, confusion, inconsistencies, and architecture built on faulty premises. The author emphasizes that this shift is less applicable to large or existing applications and team-based work. ### Source excerpt Managing comprehension debt when leaning on AI to code ## How to write a good spec for AI agents DevFeed: [How to write a good spec for AI agents]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-01-19T15:31:16Z Content type: tutorial Language: en Sources: [Elevate]() Topics: [AI-assisted coding](), [AI Agent](), [Code](), [context window](), [Claude Code]() Tags: [agents](), [claude-code](), [cli](), [coding](), [coding-agents](), [context-window](), [how-to](), [prompting]() ### AI overview A practical guide to writing specifications for AI coding agents. It recommends starting with a concise high-level vision, adding enough structure, style, testing, and boundaries to guide the agent, breaking large tasks into smaller ones, and iterating on a plan before implementation. ### Source excerpt How to structure, plan, and iterate for high-performance coding agents ## Code Review in the Age of AI DevFeed: [Code Review in the Age of AI]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2026-01-05T15:30:39Z Content type: article Language: en Sources: [Elevate]() Topics: [Artificial Intelligence](), [Code review](), [AI-assisted coding](), [Testing](), [Pull Request](), [GitHub Copilot code review]() Tags: [agents](), [ai](), [code-review](), [github-copilot](), [pull-request](), [security](), [testing]() ### AI overview The article argues that AI has not eliminated code review; it has made verification and evidence more important. Developers should support changes with manual checks and automated tests, while using review to assess risk, intent, context, and accountability. ### Source excerpt AI writes faster. Humans still have to prove it works. ## How AI Performs at React Coding and How Developers Can Improve Its Results DevFeed: [How AI Performs at React Coding and How Developers Can Improve Its Results]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2025-12-29T15:31:06Z Content type: tutorial Language: en Sources: [Elevate]() Topics: [React](), [Artificial Intelligence](), [ai-coding](), [coding](), [Development]() Tags: [ai](), [ai-assistants](), [ai-coding](), [benchmarks](), [coding](), [context-engineering](), [guide](), [react]() ### AI overview This article examines how AI coding tools perform for React developers. It reports stronger results on isolated tasks such as scaffolding components and implementing explicit requirements, but weaker results on multi-step integrations involving complex state management and design judgment. It argues that context engineering, specific prompts, structured workflows, guardrails, and deep React knowledge can improve outcomes. ### Source excerpt A data-driven look at what AI can and can't do for React developers - and what you can do about it ## My LLM coding workflow going into 2026 DevFeed: [My LLM coding workflow going into 2026]() Original publisher: [Read original article]() Author: Addy Osmani Published: 2025-12-18T15:30:46Z Content type: tutorial Language: en Sources: [Elevate]() Topics: [AI-assisted coding](), [Large Language Model](), [Requirements]() Tags: [ai-coding](), [best-practices](), [llm](), [requirements]() ### AI overview The article presents an AI-assisted coding workflow that treats an LLM as a pair programmer requiring direction, context, and human oversight. It recommends defining requirements and edge cases, compiling a specification, and iterating on a project plan before generating code. ### Source excerpt Best practices for staying in control while coding with AI