Best of Code Review — February 2026

  1. 1
    Article
    Avatar of addyAddy Osmani·15w

    Agentic Engineering

    Agentic engineering is a disciplined approach to AI-assisted software development that distinguishes itself from "vibe coding" through human oversight and engineering rigor. While vibe coding means accepting AI output without review (useful for prototypes and MVPs), agentic engineering involves treating AI agents as tools that handle implementation under careful human direction. The workflow requires writing specs before prompting, reviewing every diff, running comprehensive test suites, and maintaining ownership of the codebase. This approach disproportionately benefits senior engineers with strong fundamentals, as it trades typing time for review time and demands architectural thinking over raw code generation. The rise of AI coding raises rather than lowers the bar for software engineering craft.

  2. 2
    Article
    Avatar of nodelandAdventures in Nodeland·13w

    Yes, Learning to Code Is Still Valuable

    AI hasn't eliminated the need to learn coding—it's shifted the bottleneck from writing code to reviewing it. You can't develop good technical judgment without first understanding how software works through hands-on experience. While AI can generate code faster, humans still need deep knowledge of fundamentals like algorithms, distributed systems, and data structures to evaluate what AI produces. The traditional bootcamp path may be closing, but learning to code remains essential for anyone who needs to maintain, debug, or make architectural decisions about software systems.

  3. 3
    Article
    Avatar of techworld-with-milanTech World With Milan·12w

    What I learned from the book Software Engineering at Google

    A detailed breakdown of key lessons from the book 'Software Engineering at Google', covering the distinction between programming and engineering, Hyrum's Law, the Beyoncé Rule, shift-left testing, why mocking frameworks are discouraged in favor of fakes, code review best practices, small frequent releases, dependency management, the GSM productivity framework, and engineering culture. The post also includes honest admissions from the authors about what doesn't work even at Google, and closes with practical takeaways applicable to teams of any size.

  4. 4
    Article
    Avatar of systemdesignnewsSystem Design Newsletter·15w

    I struggled to code with AI until I learned this workflow

    AI coding assistants work best through an iterative workflow rather than one-shot prompts. The key is providing comprehensive context (project background, constraints, relevant code), requesting a plan before implementation, generating code in small steps with defined roles (planner, implementer, tester, explainer), reviewing output with AI-assisted tools, writing tests immediately, and debugging systematically. Common pitfalls include context drift in long conversations, API version mismatches, and over-reliance on AI without understanding the output. The workflow emphasizes treating AI like a new teammate who needs explicit briefing, keeping changes small and reviewable, and maintaining human oversight throughout the process.

  5. 5
    Article
    Avatar of colkgirlCode Like A Girl·14w

    You Don’t Suck at Coding. You’re Just Meeting a Real Codebase.

    Transitioning from college to corporate software engineering means encountering large, complex codebases built under real-world constraints like deadlines, legacy systems, and business pressures. The initial confusion isn't a sign of incompetence but a natural response to code archaeology—understanding decisions made by previous developers under specific circumstances. Success comes from reading and understanding existing code rather than writing perfect new code, building familiarity incrementally, and recognizing that senior engineers also face uncertainty but approach it with calm confidence gained through accumulated experience.

  6. 6
    Article
    Avatar of devtoDEV·15w

    Above the API: What Developers Contribute When AI Can Code

    AI coding assistants create a divide between developers who use them for delegation versus judgment. Research shows junior engineers using AI finish faster but score 17% lower on mastery tests. The critical skills that remain valuable are architectural thinking, verification capability, maintenance of existing systems (v2+), simplification discipline, and domain expertise. These "above the API" skills are traditionally learned through friction, mentorship, and public knowledge sharing—transmission mechanisms now at risk. Developers who treat AI as a confident junior requiring review maintain value, while those who blindly accept AI output lose understanding. The piece argues for deliberate verification habits, public knowledge contribution, and explicit mentorship to preserve these judgment skills across generations.

  7. 7
    Article
    Avatar of programmingdigestProgramming Digest·13w

    The PERFECT Code Review: How to Reduce Cognitive Load While Improving Quality – Daniil Bastrich

    The PERFECT framework provides a structured approach to code reviews, prioritizing Purpose, Edge cases, Reliability, Form, Evidence, Clarity, and Taste in descending order of importance. The methodology aims to reduce cognitive load and procrastination by establishing clear conventions, requiring self-review before peer review, and focusing on objective criteria over subjective opinions. Implementation recommendations include maintaining written conventions, automating checks, integrating review into development workflows, and treating code review as a skill that improves with practice.

  8. 8
    Video
    Avatar of philipplacknerPhilipp Lackner·13w

    Learning THIS Becomes More Important Than Ever In the Era of AI

    As AI becomes better at writing boilerplate code, solving well-defined problems, and reviewing logic, developers need to shift from a 'bricklayer' mindset to an 'entrepreneurial' one. The skills that matter most going forward are system design and architecture (which require deep organizational context AI can't replicate), deep technical understanding for reviewing AI-generated code, and hands-on experience actually using AI tools and agents in practice. Mobile developers in particular are encouraged to start experimenting with AI in their IDEs, refine their prompting skills, and treat AI as a fast execution layer they supervise rather than a replacement for engineering judgment.

  9. 9
    Article
    Avatar of cursorCursor·13w

    How Stripe rolled out a consistent Cursor experience for 3,000 engineers · Cursor

    Stripe rolled out Cursor to 3,000+ engineers by preinstalling it on every machine, using Cursor Rules for codebase context, and adapting code review practices. They found that senior engineers with deep codebase knowledge gained the most productivity, contrary to expectations that juniors would benefit most. The company maintained quality by using LLMs to flag risky code during reviews and spread adoption through power users sharing workflows.

  10. 10
    Article
    Avatar of planetpythonPlanet Python·15w

    Coding can be super lonely

    Solo coding can lead to isolation and stagnation without peers to share challenges, review code, or provide feedback. Working in a vacuum reinforces bad habits and limits growth. Breaking this cycle requires seeking code reviews focused on logic, building tools for others to create accountability, and engaging with developer communities. Growth comes from friction with other developers and community engagement, not from coding alone.

  11. 11
    Article
    Avatar of minersThe Miners·12w

    The Double Standard Is Killing AI Adoption in Your Team

    Developers apply a double standard when reviewing AI-generated code, demanding perfection from agents while routinely approving untested, poorly structured human-written code. Drawing on Linus Torvalds' 1992 defense of Linux against Tanenbaum's microkernel critique and Richard Gabriel's 'Worse is Better' essay, the argument is that shipping functional, tested code has always mattered more than theoretical elegance. AI-generated code that compiles, runs, and includes tests deserves the same pragmatic review standard applied to human code — not a higher bar.

  12. 12
    Article
    Avatar of phoronixPhoronix·14w

    Linus Torvalds Rejects MMC Changes For Linux 7.0 Cycle: "Complete Garbage"

    Linus Torvalds rejected the MMC subsystem pull request for Linux 7.0, calling the changes "complete garbage" that don't compile. The code failed basic build testing when CONFIG_MULTIPLEXER=m due to incorrect preprocessor conditionals, causing redefinition errors. Torvalds refused further pull requests this merge window and demanded the code be properly tested in linux-next before resubmission in 7.1.

  13. 13
    Article
    Avatar of hnHacker News·14w

    Stop generating, start thinking

    LLM-generated code is fundamentally different from traditional automation because it's non-deterministic and cannot reason about system architecture or context. While tools like Copilot can serve as autocomplete, outsourcing actual thinking to AI removes accountability and shared understanding from software development. The practice mirrors fast fashion: quick but low-quality output trained on existing poor code. Unlike higher-level programming languages that abstract complexity predictably, LLMs eliminate the critical thinking that ensures maintainable, accessible, and safe software. The author advocates keeping humans in the loop for reasoning while using AI only for tasks you already understand.

See all Code Review archives