Best of Code Review — January 2026

  1. 1
    Article
    Avatar of nodelandAdventures in Nodeland·17w

    The Human in the Loop

    AI has fundamentally changed software development workflows, but human review and judgment remain critical. While AI can generate code rapidly for bug fixes and features, engineers must review every change to ensure correctness, security, and architectural fit. The bottleneck has shifted from coding speed to review capability. The role of programmers who simply execute tasks is obsolete, but software engineers and architects who provide judgment, understand context, and maintain accountability are more crucial than ever. The real risk isn't AI replacing developers, but creating a culture where shipping unreviewed AI-generated code becomes acceptable.

  2. 2
    Article
    Avatar of addyAddy Osmani·19w

    My LLM coding workflow going into 2026

    A comprehensive workflow for using LLM coding assistants effectively in 2026. Start with detailed planning and specs before coding, break work into small iterative chunks, provide extensive context to the AI, choose appropriate models for each task, and maintain human oversight through rigorous testing and code review. Use version control aggressively with frequent commits, customize AI behavior with rules and examples, leverage automation as quality gates, and treat the AI as a powerful but fallible pair programmer requiring clear direction. The approach emphasizes that AI amplifies engineering skills rather than replacing them, with the developer remaining accountable for all code produced.

  3. 3
    Article
    Avatar of planeterlangPlanet Erlang·19w

    Software Acceleration and Desynchronization

    Software development acceleration creates desynchronization across interconnected work loops. When teams speed up individual tasks like code writing, they risk decoupling from slower but essential feedback cycles around operations, architecture, and organizational knowledge. This desynchronization accumulates as drift between mental models and reality, potentially leading to incidents that force rapid resynchronization. Strategic slowdowns in certain areas can actually accelerate overall system performance by maintaining necessary synchronization points. The drive for continuous acceleration is a self-reinforcing temporal structure that shapes how software organizations function, requiring careful analysis of which loops to speed up and which provide essential stability.

  4. 4
    Article
    Avatar of avanderleeSwiftLee·17w

    11 Things I learned after using AI Agents full-time

    AI agents like Cursor have transformed app development workflows through deliberate practices. Key learnings include using plan mode before execution to clarify intent, maintaining custom GPT projects for context continuity, leveraging agent skills instead of lengthy AGENTS.md files, choosing reasoning models for complex tasks, breaking large prompts into focused scopes, reviewing AI-generated code carefully to prevent tech debt, learning tools deeply to find leverage points, implementing linters and rules as guardrails, evolving AGENTS files to prevent recurring mistakes, running tests and CI on every PR, and exploring uncomfortable tools to discover new capabilities. The core insight: AI agents amplify both discipline and shortcuts, multiplying responsibility rather than removing it.

  5. 5
    Article
    Avatar of rubylaRUBYLAND·17w

    How I Read A Pull Request

    A systematic approach to reviewing pull requests using the journalist's 5 Ws and H framework. The process covers when to review (start of day, between meetings, after submitting your own PR), what to look for based on PR intent, understanding your motivation as a reviewer, considering the author and audience, and a specific reading order: title/description first, then tests (test-driven review), followed by implementation alphabetically, and finally a big-picture pass to verify goals and overall structure.

  6. 6
    Article
    Avatar of addyAddy Osmani·19w

    AI writes code faster. Your job is still to prove it works.

    AI-generated code is becoming mainstream, with over 30% of senior developers shipping mostly AI-written code by early 2026. However, AI-generated code contains 75% more logic errors and 45% security flaws compared to human code. The key shift is that code review now focuses on verification and accountability rather than line-by-line inspection. Solo developers can ship at "inference speed" by relying on comprehensive automated testing, while teams must maintain human oversight for security, context sharing, and maintainability. The bottleneck has moved from writing code to proving it works through evidence like tests, manual verification, and clear documentation of AI's role.

  7. 7
    Article
    Avatar of mitsuhikoArmin Ronacher·17w

    Agent Psychosis: Are We Going Insane?

    AI coding agents are creating a productivity paradox where developers experience dopamine-driven addiction to agent-assisted coding while producing low-quality contributions that burden maintainers. The asymmetry between generating AI code (minutes) and reviewing it (hours) is creating friction in open source projects. Tools like Beads and Gas Town exemplify extreme cases where developers build complex systems with minimal oversight, resulting in unmaintainable codebases. While AI agents offer genuine productivity gains, uncritical reliance leads to "slop" code, token waste, and parasocial relationships with AI that distort judgment about code quality.

  8. 8
    Article
    Avatar of zedZed·16w

    On Programming with Agents — Zed's Blog

    LLMs automate typing, not thinking. When working with AI agents for coding, focus on three core principles: only use agents for tasks you already understand, stay actively involved throughout the process to catch when agents go off-track, and rigorously review all generated code. Create detailed plans upfront, keep tasks small enough to review in one sitting, and treat agent-generated code with stricter scrutiny than colleague contributions. This approach allows developers to maintain higher standards by offloading mechanical work while preserving the craft of thoughtful software design.

See all Code Review archives