JavaScript & Web — devopsmantra wiki

1,347 saved links, 2015–2026 (peak: 2023). Your biggest tag and your loosest one — a decade of watching the web platform from the ops side of the fence, 701 web pages and 566 articles deep. The older strata are free-React-course listicles and certification roundups; the 2024–2026 layer tells a very different story: you stopped saving “learn JavaScript” and started saving “let the agent write the frontend.” That turn — AI as the web developer, HTML as the agent’s deliverable, spec-driven development as the new craft — is what this page curates.

Related: Programming · AI Agents · APIs · UX Design · Learning Resources

The turn: AI builds the frontend now#

The clearest arc in your recent saves. In early 2024 it was a novelty; by 2026 it’s how frontend work gets done.

“I have zero coding skills, but I coaxed ChatGPT to write a JavaScript blogging tool” — the 2024 marker of where this thread starts.
an AI-powered web app in 20 minutes, no frontend skills required — the 2025 version.
David Bau teaches himself to vibe code — watching Claude Code grow 780 lines to 13,600, with two rules for staying in control; matklad’s vibecoding notes — how a very conscientious developer chooses where to use LLMs and where not to.
a multi-agent harness for frontend design and long-running autonomous engineering and designing delightful frontends with GPT-5.4 — both labs publishing frontend-generation playbooks.
Vercel’s react-best-practices — a repo written for coding agents, with evals to catch accidental waterfalls and bundle bloat, plus the Next.js evals and Google’s Modern Web Guidance skills — you’d already built your own semantic-HTML skill when these landed.
AI-assisted coding: a practical guide — “that gap between code that works and code that is production-ready? It’s you.” And the free Claude Code course with Anthropic — agentic loop, CLAUDE.md, skills, hooks, subagents.
teaching an LLM to review code like a senior engineer (video) — the future-of-frontend talk.

HTML is the new deliverable#

A distinctly 2026 pattern in your saves: self-contained HTML as the universal output format for agents — docs, slides, courses, diagrams.

effective-html — HTML skills for coding agents: architecture diagrams, plan pages, visual documents, all self-contained.
codebase-to-course — one instruction turns any repo into an offline interactive HTML course; from the same author, frontend-slides (“saved my life… I haven’t manually made slides in weeks”).
ht-ml.app — agent-native HTML artifact sharing with WebMCP support, and html-docs.com — your note: “holy shit, really well done.”
AI-generated educational HTML5 animations.

The martinfowler.com subscription#

The steadiest drip in the whole tag — Thoughtworks watching the same AI-eats-the-SDLC story you are, with more rigor.

harness engineering — Birgitta Böckeler’s mental model and sensors for coding agents — static analysis as the agent’s nervous system.
“What is code?” — your note: “if you think code no longer matters, read this.”
encoding team standards as infrastructure — versioned, reviewed prompt instructions instead of tribal knowledge.
the Interrogatory LLM — get the LLM to interview you for context rather than writing it yourself.
Fragments: future libraries may be specs — and LLMs as a GPS, not a destination.

Spec-driven development & the craft of agentic coding#

Drew Breunig’s blog is your second firehose, and it’s converging on one idea: prompts don’t scale, specs do.

the problem is prompt debt — “prompts are great for one-off requests, terrible for defining the behaviors of systems.”
the rise of spec-driven development and the spec-driven development triangle — logging decisions made by you and the agent, structured and tracked.
10 lessons for agentic coding · system prompts across six coding agents, compared · how Claude Code builds its system prompt.
The context trilogy: How Long Contexts Fail · How to Fix Your Context · Let the Model Write the Prompt.
Armin Ronacher on “shitty types” — why TypeScript can both help and harm agentic coding, and why Go doesn’t suffer the same.
the new calculus of AI-based coding — productive at speed, but knowing what breaks at speed.

Where JavaScript meets your AI stack#

The practitioner corner — the saves where web tech plugs into the agent/MCP/Azure world you actually build in.

use-mcp — connect React apps to any MCP server in 3 lines, plus Cloudflare’s launch post and their JavaScript framework for building agents.
CoAgents — LangGraph agents inside React apps with generative UI and human-in-the-loop, and streaming Google ADK agents into modern frontends via the AG-UI protocol.
generative-ai-with-javascript — learn AI while chatting with da Vinci and Ada Lovelace, and the companion video series.
The Azure/Node.js shelf: Cosmos DB + Azure OpenAI Node.js developer guide — 13 modules · DeepSeek with Azure AI in JavaScript/TypeScript · a travel AI agent: React + FastAPI + Cosmos vector store.
microblog-ai-nextjs and its zero-to-AI-hero Next.js workshop — the fullest end-to-end build you’ve saved.
the Assistants API + Next.js quickstart · mcpservers.org — thousands of MCP servers and skills, front-end category included.

Fundamentals worth keeping#

The evergreen minority in a newsy tag — the things that stay true whoever writes the code.