Free · No email required

The coding roadmap I wish someone gave me at 14

Most roadmaps show you a map of everything and let you drown in it. This one is opinionated. It tells you what to pick, what order to learn it in, and — the part nobody marks — exactly where the job-ready line sits.

Read this before anything else

Rule 1 — Pick one language. Go stupidly deep.

This isn't my opinion, it's what the major curricula actually say. roadmap.sh's backend track states it plainly: "Learn one language and build lots of projects before moving on." The Odin Project is blunter: "The important thing is to not switch halfway through; this will only set you back."

The people still stuck after two years usually didn't pick wrong. They kept starting over.

Rule 2 — Python or JavaScript. The choice is about your goal, not the language.

No credible source claims one is objectively the better first language. What actually decides it:

JavaScript if you want to build for the web. You'll have to learn it eventually anyway — it's the only language browsers run. Python if you want AI, data, or automation. It's also the prerequisite language of the whole AI-engineering world.

Rule 3 — Be careful how you use AI while you're learning.

Anthropic ran a randomised controlled trial with 52 mostly-junior engineers learning an unfamiliar Python library. The group using an AI assistant scored 50% on the comprehension quiz. The group coding by hand scored 67% — close to two letter grades. The biggest gap was in debugging. And they were only ~2 minutes faster, which wasn't statistically significant.

The people who did well used AI to explain things, not to produce things. So: try it yourself first, get stuck, then ask AI why. Don't skip the getting-stuck part — that's where the learning actually happens.

Pick your path

Frontend Web →

You want to build things people can see and click. Fastest visual feedback.

Backend →

You want to build the systems behind the app. Data, APIs, servers.

Full-Stack →

Both. Longer, but you can build an entire product alone.

AI Engineer →

Building products on top of AI models. Not a beginner path — read the warning.

Frontend Web

Note the name: Frontend Web. This path is about browsers. Mobile and desktop are different roads.

  1. How the internet worksHTTP, DNS, domains, hosting, what a browser actually does. One afternoon. Don't skip it — everything later assumes it.
  2. HTMLStructure. Semantic tags, forms, accessibility basics.
  3. CSSLayout, flexbox, grid, responsive design. This is where most people rush and regret it.
  4. JavaScriptThe actual language. Not a framework. Functions, arrays, objects, the DOM, fetch, async/await.
Checkpoint — build 3 ugly projects with no framework. A landing page, a calculator, something that fetches from a public API. Ugly is fine. Finished is the point.
  1. Git & GitHubVersion control, branches, pull requests. Push everything from here on.
  2. Package managersnpm, or pnpm/Bun. Understand what node_modules is.
  3. A CSS frameworkTailwind is the default now.
  4. One JavaScript frameworkReact, Vue, Svelte — pick one. React has the most jobs. Only after vanilla JS is solid.
  5. Using AI in developmentNew as core curriculum in 2026, and it sits here — after fundamentals, never before. Assisted coding, prompting, code review, refactoring.
  6. TypeScriptTypes on top of JavaScript. Effectively required for jobs now.
  7. DeploymentVercel, Netlify, or Cloudflare. Get things live on the internet.
← The job-ready line is roughly here.

With solid HTML/CSS/JS, one framework, TypeScript, Git and deployed projects, you can apply. Everything below makes you better, not employable-er.

Optional / after your first job
  1. Build toolsVite, esbuild, Rollup.
  2. TestingVitest, Playwright, Cypress.
  3. Web securityHTTPS, CORS, CSP, the OWASP basics.
  4. Performance & accessibilityLighthouse, caching, service workers, screen readers.
  5. SSR / SSGNext.js, Astro.
  6. Design systems, GraphQL, PWAs, mobile/desktopReact Native, Flutter, Electron, Tauri.

Backend

The systems behind the screen. Data, logic, APIs.

  1. How the internet worksSame starting point. HTTP, DNS, hosting.
  2. Basic HTML & CSSEnough to understand what the frontend sends you. A weekend.
  3. Pick ONE languageJavaScript/Node, Python, Go, Java, C#, Ruby, PHP, Rust. All of them get you hired. Choosing is the hard part; unchoosing repeatedly is what kills people.
Checkpoint — build several projects in that one language before you touch anything else on this list.
  1. Git & GitHubNon-negotiable.
  2. Relational databasesPostgreSQL is the safe default. Schemas, joins, indexes, migrations, the N+1 problem.
  3. APIsREST first. JSON, status codes, OpenAPI. GraphQL and gRPC can wait.
← The job-ready line is here, and it's earlier than you think.

roadmap.sh marks it right after APIs — before Docker, Kubernetes, microservices and message queues. Most beginners believe they need the whole map first. They don't.

Optional / after your first job
  1. CachingRedis, HTTP caching.
  2. AuthenticationJWT, OAuth, sessions, password hashing (bcrypt).
  3. Web securityHTTPS, TLS, CORS, CSP, OWASP Top 10.
  4. Using AI in developmentPlus building AI features: streaming, structured outputs, function calling, embeddings, RAG.
  5. Testing & CI/CDUnit, integration, GitHub Actions.
  6. DockerThen Kubernetes only if you actually need it.
  7. Message brokers, architecture patterns, scalingKafka, RabbitMQ, microservices, replication, sharding, CAP.

Full-Stack

Frontend plus backend. This is the only path with checkpoints baked into the official roadmap, and they're worth following literally.

  1. HTML + CSS→ Checkpoint: static webpages
  2. JavaScript→ Checkpoint: interactivity
  3. npm→ Checkpoint: using external packages
  4. Git + GitHub→ Checkpoint: collaborative work
  5. React + Tailwind→ Checkpoint: frontend apps
  6. Node.js→ Checkpoint: CLI apps. Recommended because you already know JavaScript — one less language to context-switch into.
  7. PostgreSQL→ Checkpoint: simple CRUD apps
  8. REST APIs + JWT auth + Redis→ Checkpoint: a complete application
← Job-ready around here.

You can build and ship a whole product by yourself. That's the bar.

Optional / after your first job
  1. Linux basics + a cloud provider→ Checkpoint: deployment
  2. Monitoring→ Checkpoint: you know when it breaks
  3. GitHub Actions→ Checkpoint: CI/CD
  4. Ansible, Terraform→ Checkpoint: automation & infrastructure

On the "3 months to full-stack" claim

The official roadmap says 3 months is realistic only if you already know one half and are filling in the other. From zero, it isn't.

AI Engineer

This is not a beginner path

The official roadmap lists a hard prerequisite: frontend, backend, or full-stack first. AI engineering is a specialisation you add on top of being a developer, not a shortcut around becoming one.

Also worth separating: an AI Engineer builds products using existing models. An ML Engineer builds and trains the models — that's the path with the linear algebra and statistics. This isn't that.

  1. Be a developer firstAny of the three paths above. Non-negotiable.
  2. TerminologyLLMs, tokens, embeddings, context windows, inference, RAG, agents, fine-tuning.
  3. How LLMs actually workTemperature, top-k, top-p, sampling, repetition penalties.
  4. Prompt engineeringZero-shot, few-shot, chain-of-thought, ReAct.
  5. Context engineeringNew in 2026 and now its own discipline. System prompting, context compaction, isolation, external memory.
  6. Working with modelsStructured output, function calling, prompt caching, streaming. Providers and APIs.
  7. Embeddings & vector databasespgvector, Chroma, Pinecone, Qdrant, FAISS.
  8. RAGChunk → embed → index → retrieve → generate. And knowing when to use RAG instead of fine-tuning.
  9. AI agents & MCPTool calling, multi-agent systems, the Model Context Protocol.
  10. Safety & ethicsPrompt injection, bias, moderation, adversarial testing.
  11. ObservabilityTracing, cost and latency monitoring. LangSmith, Langfuse, Helicone.
  12. EvaluationsDeterministic evals, model-as-judge, human evals, regression testing. The rarest skill on this list — most people building with AI have no idea whether their last prompt change made things worse.

How long does this actually take?

Every "become a developer in 6 months" claim I could find traced back to a company selling a course. None published a methodology. So here are only the numbers that come from the free curricula themselves — divide by your own hours per week.

SourceWhat they state
freeCodeCamp~300 hours per checkpoint certification; ~1,800 hours for the full Certified Full Stack Developer capstone
Full Stack Open15–20 hours per part, 14 parts. Requires 100–200 hours of prior programming experience first
Teach Yourself CS100–200 hours per subject, 9 subjects
roadmap.sh (full-stack)3 months — but only if you already know one half

The more useful answer

Stop counting months. The backend roadmap marks "you know enough to get a job" immediately after learning APIs — long before the scary infrastructure topics. Scope beats time as a measure of progress.

The rules that matter more than the list

1. Pick one path. Finish it. Switching is the single most common way people lose two years.

2. Build ugly things. A finished ugly project teaches more than a beautiful tutorial you followed along with.

3. Learn the language before the framework. React on top of shaky JavaScript collapses the moment something breaks.

4. Get stuck on purpose. Try it yourself, fail, then ask AI to explain. The struggle is the mechanism, not an obstacle to it.

5. Ship publicly from day one. Push to GitHub, deploy it, post about it. Proof compounds; private work doesn't.

6. You need far less than you think to get hired. The job-ready line is much earlier on every one of these maps than beginners assume.