Best of AI — December 2025

  1. 1
    Article
    Avatar of cassidooCassidy's blog·22w

    Vibe coding is boring

    Vibe coding with AI agents is effective for shipping side projects quickly, but removes the satisfaction and learning that comes from hands-on development. While tools like GitHub Copilot and Spec Kit can automate implementation from specifications, watching agents write code is tedious and lacks the joy of problem-solving. The author reserves AI-assisted coding for projects where only the final output matters, preferring to manually build applications where the tech stack or implementation details are interesting.

  2. 2
    Article
    Avatar of collectionsCollections·23w

    Anthropic Acquires Bun to Enhance AI Coding Products' Performance and Stability

    Anthropic acquired Bun, the high-performance JavaScript runtime with 7 million monthly downloads and 83,000 GitHub stars. Despite $26 million in VC funding, Bun generated no revenue. The runtime will remain open source under MIT license and integrate into Anthropic's AI coding products like Claude Code and Claude Agent SDK. The acquisition provides Bun with resources to expand its team and accelerate development while supporting Anthropic's AI-driven development infrastructure, as Claude Code reaches $1 billion in run-rate revenue.

  3. 3
    Article
    Avatar of itsfossIt's Foss·20w

    F*** You! Co-Creator of Go Language is Rightly Furious Over This Appreciation Email

    Rob Pike, co-creator of the Go programming language and legendary computer scientist from Bell Labs, expressed outrage after receiving an AI-generated thank-you email. The email came from an AI agent participating in the AI Village project, where agents were tasked with performing "random acts of kindness" and interpreted this by sending unsolicited emails to famous programmers. Pike's angry response highlights concerns about AI-generated content wasting resources, the environmental cost of AI infrastructure, and the broader societal impact of meaningless AI-generated material flooding our digital spaces.

  4. 4
    Article
    Avatar of elevateElevate·21w

    My LLM coding workflow going into 2026

    A comprehensive guide to using LLM coding assistants effectively in 2026. Key practices include starting with detailed specifications before coding, breaking work into small iterative chunks, providing extensive context to the AI, choosing appropriate models for different tasks, maintaining human oversight through testing and code review, committing frequently for version control safety, customizing AI behavior with rules and examples, leveraging automation as quality gates, and treating AI as a force multiplier rather than replacement. The workflow emphasizes treating LLMs as junior pair programmers requiring guidance while maintaining developer accountability for all code produced.

  5. 5
    Article
    Avatar of weprodevWeProDev·19w

    Learn Python in 2026.

    Python remains essential in 2026, particularly for AI workflows. The best approach is learning by writing and running code rather than passive consumption. A structured, example-driven GitHub repository provides a roadmap from fundamentals to advanced topics. Python skills compound over time, enabling better understanding of AI ecosystems, faster code adaptation, and effective debugging of integrations. Recommended learning resources include Codebasics and Corey Schafer YouTube channels for practical, engineering-focused tutorials.

  6. 6
    Article
    Avatar of bradfrostBrad Frost·21w

    Agentic Design Systems in 2026

    Design systems combined with AI enable new collaborative workflows where non-technical team members can verbally describe features and see them generated using production-grade components. The Storybook MCP tool demonstrates this approach, constraining AI generation to use established design system standards rather than arbitrary code generation. This "DS+AI" formula creates coded prototypes that short-circuit traditional design review cycles while maintaining organizational quality standards.

  7. 7
    Article
    Avatar of freecodecampfreeCodeCamp·22w

    Learn n8n to Design, Develop, and Deploy Production-Grade AI Agents

    n8n is an open-source visual workflow automation tool for connecting applications, APIs, and AI models. A comprehensive beginner course covers building practical AI agents including email automation, research workflows with OpenAI and Perplexity, and a customer support RAG agent using vector databases like Pinecone. The training includes advanced topics like modular component patterns, multi-workflow builds for coordinating agent teams, and deployment options including cloud, Docker, and self-hosting with local LLMs like Ollama.

  8. 8
    Article
    Avatar of langchainLangChain·22w

    Agent Engineering: A New Discipline

    Agent engineering is an iterative discipline for building reliable LLM-based agents in production. It combines product thinking (prompt writing, defining scope), engineering (building tools, infrastructure, UI), and data science (evaluation, monitoring, analysis) in a continuous cycle of build, test, ship, observe, and refine. Unlike traditional software, agents handle unpredictable natural language inputs and non-deterministic behavior, making production deployment essential for learning what actually works. Successful teams treat shipping as a learning mechanism rather than an end goal, using tracing and evaluation to systematically improve agent reliability through rapid iteration.

  9. 9
    Article
    Avatar of itsfossIt's Foss·22w

    No AI Slops! GNOME Now Forbids Vibe Coded Extensions

    GNOME has updated its extension review guidelines to reject AI-generated code submissions. The policy targets low-quality extensions with unnecessary code patterns like excessive try-catch blocks, inconsistent styling, and imaginary API usage. Reviewers were spending over 6 hours daily reviewing 15,000+ lines of code, much of it AI-generated slop. Using AI as a learning tool or for code completions remains allowed; the ban specifically targets developers who generate entire extensions without understanding the code.

  10. 10
    Article
    Avatar of dailydoseofdsDaily Dose of Data Science | Avi Chawla | Substack·20w

    The AI Engineering Guidebook

    A comprehensive 350+ page guidebook covering the engineering fundamentals of LLM systems, including model architecture, training, prompt engineering, RAG systems, fine-tuning techniques like LoRA, AI agents, Model Context Protocol, optimization strategies, and deployment considerations. The resource focuses on practical engineering decisions, system design tradeoffs, and real-world implementation patterns rather than surface-level usage.

  11. 11
    Article
    Avatar of webcraftWebCraft·21w

    No-Code Is Dying And Honestly, It’s About Time.

    No-code platforms are evolving from simple drag-and-drop tools into sophisticated development engines that complement rather than replace traditional coding. Modern no-code tools now offer custom databases, API integrations, advanced workflows, and extensibility through code blocks, making them valuable productivity multipliers for developers building internal tools, prototypes, and automation layers. The integration of AI assistants is accelerating this transformation, enabling both technical and non-technical users to build faster while blurring the lines between no-code and low-code development.

  12. 12
    Article
    Avatar of linearLinear·21w

    Design is more than code

    Design should focus on understanding and defining problems before jumping to solutions, rather than being reduced to code execution. The design process involves two stages: conceptual (finding the right form and direction based on problem understanding and product vision) and execution (building it out). While new tools and AI make execution easier, there's a risk of devaluing the strategic thinking that happens before coding—questioning problems, aligning stakeholders, and making intentional decisions about product direction. The concern isn't about whether designers should code, but whether the industry will lose the patience for deep consideration and problem-solving in favor of rapid output.

  13. 13
    Article
    Avatar of nolanlawsonRead the Tea Leaves·21w

    How I use AI agents to write code

    A developer shares practical strategies for using AI coding agents effectively after transitioning from skepticism to adoption. Key recommendations include creating comprehensive CLAUDE.md files for project context, using automated tests as feedback loops, running separate AI sessions for code review to catch bugs, and leveraging agents for overnight work on side projects. The author acknowledges AI's limitations with UI work and novel projects, describes the shift toward an architect-like role focused on specs and review, but maintains reservations about using AI for open-source contributions due to ownership concerns.

  14. 14
    Article
    Avatar of stackovStack Overflow Blog·19w

    A new era of Stack Overflow

    Stack Overflow announces a major rebrand and strategic shift to become "the world's most vital source for technologists" in the AI era. The company introduces a new mission focused on cultivating community, powering learning, and unlocking growth. Key updates include simplified brand architecture (Stack Overflow for public platform, Stack Overflow Business for enterprise), new engagement features like Community Activity and enhanced Chat, an AI-powered search tool (stackoverflow.ai), Coding Challenges, and enterprise capabilities including Knowledge Ingestion and new integrations with Microsoft Graph, Backstage.io, and Moveworks. The announcement emphasizes addressing the growing trust gap in AI tools, with 46% of developers not trusting AI output despite 83% using AI tools.

  15. 15
    Article
    Avatar of tcTechCrunch·21w

    Google's vibe-coding tool Opal comes to Gemini

    Google is integrating Opal, its vibe-coding tool for building AI-powered mini apps, directly into the Gemini web interface. Users can now create custom apps (called Gems) using natural language descriptions, with a visual editor that arranges steps without writing code. The tool includes a new view that converts written prompts into step-by-step workflows, and advanced users can access more customization options at opal.google.com. This move positions Google alongside other AI-powered app-building tools like Lovable, Cursor, and offerings from Anthropic and OpenAI.

  16. 16
    Article
    Avatar of devtoDEV·22w

    The Vibe Coding Paradox

    AI coding assistants amplify existing system patterns rather than improving them. When working on well-architected personal projects with clear ownership, AI extends intentional design decisions and maintains quality standards. In legacy contract codebases with technical debt, the same tools make "good enough" solutions trivially easy, perpetuating dysfunction with clean-looking code. The friction that once forced developers to decide what mattered has been removed, shifting the craft from typing to judgment about when to care and when to ship.

  17. 17
    Article
    Avatar of hnHacker News·21w

    AWS CEO Explains 3 Reasons AI Can’t Replace Junior Devs

    AWS CEO Matt Garman argues against replacing junior developers with AI, citing three key reasons: junior developers are often more proficient with AI tools than senior staff, they represent minimal cost savings as the lowest-paid employees, and eliminating them breaks the talent pipeline that companies need for future growth. He emphasizes that junior hires bring fresh perspectives, adapt quickly to new technologies, and form the foundation for developing future leaders. While acknowledging AI will change how developers work, Garman believes it will ultimately create more jobs than it eliminates in the medium to long term.

  18. 18
    Article
    Avatar of engineerscodexEngineer’s Codex·21w

    Everyone is a Staff Engineer Now

    AI coding agents like Claude Code are shifting engineering work from implementation to higher-level activities like architecture, planning, and code review. Skills traditionally associated with staff engineers—maintaining context across systems, managing asynchronous workflows, steering AI agents effectively, and reviewing code—are becoming baseline requirements earlier in careers. Junior engineers now operate at senior-level abstraction while seniors architect at staff-level scale. Success depends less on prompting AI and more on managing personal context, maintaining focus during agent runtime, and developing new workflows that treat AI as a junior engineer to delegate to.

  19. 19
    Article
    Avatar of hnHacker News·20w

    Bye Bye Big Tech: How I Migrated to an almost All-EU Stack (and saved 500€ per year)

    A developer shares their complete migration from US-based tech services to EU-hosted alternatives, detailing specific tool replacements across email, cloud storage, password management, AI, hosting, and productivity. The Proton ecosystem replaced Google Workspace, 1Password, and Notion, while Scaleway replaced AWS/Azure, and Mammouth provided multi-model AI access. The migration resulted in €528 annual savings (from €83/month to €39/month) while improving privacy and data sovereignty. Challenges include limited alternatives for social platforms, blogging (still using Substack), and occasional reliance on Google search and Claude Code.

  20. 20
    Article
    Avatar of googledevsGoogle Developers·21w

    Introducing Agent Development Kit for TypeScript: Build AI Agents with the Power of a Code-First Approach

    Google released Agent Development Kit (ADK) for TypeScript, an open-source framework for building AI agents and multi-agent systems. The code-first approach lets developers define agent logic, tools, and orchestration directly in TypeScript, applying traditional software development practices like version control and CI/CD. While optimized for Gemini and Vertex AI, ADK is model-agnostic and supports third-party tools. It includes native integration with MCP Toolbox for database connections and supports the latest Gemini 3 models.

  21. 21
    Article
    Avatar of phProduct Hunt·19w

    Gitdocs AI: Make your repository explain itself.

    Gitdocs AI is a tool that automatically generates production-ready README files for GitHub repositories. It analyzes repository code and creates structured documentation with customizable templates. Recent updates include improved AI workflow, multiple documentation templates, and near-zero downtime. The service is currently free and aims to make industry-standard documentation accessible to all developers.

  22. 22
    Article
    Avatar of vigetViget·20w

    Don't Let Vibe Code Become Legacy Code

    AI-generated code can quickly become unmaintainable legacy code if not properly managed. Four key practices help prevent this: take ownership by writing code yourself or defining tests/interfaces manually, review your own pull requests thoroughly, write meaningful commit messages that explain the 'why' behind changes, and use AGENTS.md to centralize AI tooling rules across your team. These traditional software engineering best practices remain essential even when using AI coding assistants like Cursor or Claude Code.

  23. 23
    Article
    Avatar of github_updatesGitHub Changelog·21w

    Copilot Memory early access for Pro and Pro+

    GitHub Copilot Memory is now in early access for Pro and Pro+ subscribers. This feature allows Copilot agents to learn from user feedback and actions, building repository-specific memory to improve assistance across coding and code review workflows. Users can enable it through Settings > Copilot, and GitHub plans to expand availability to more subscription tiers in the future.

  24. 24
    Video
    Avatar of bigboxswebigboxSWE·20w

    VIBE CODERS MUST BE STOPPED

    Vibe coding (using AI to generate code without understanding programming) has spawned a problematic ecosystem of grifters selling expensive courses on prompting, misleading newcomers about what programming actually is. While AI coding tools have legitimate uses for experimentation or beginners, marketing them as a replacement for learning fundamentals is harmful. The practice lacks the problem-solving essence of real programming, yet practitioners often display unwarranted confidence about replacing developers. Deep technical knowledge remains valuable and irreplaceable.

  25. 25
    Article
    Avatar of meilisearchMeilisearch·23w

    I’m a developer who vibe codes – and you should, too

    Vibe coding uses AI to generate code from natural language prompts, creating an addictive, fast-paced development experience. The CEO of Meilisearch shares how this approach reignited his passion for coding despite time constraints, describing the dopamine-driven cycle of rapid iteration and instant results. While the speed and creative flow are exhilarating, challenges include mental fatigue, reduced code comprehension, and generic output. The key is balancing AI automation with human judgment, using AI for repetitive tasks while maintaining creative direction and understanding of the underlying systems.

See all AI archives