The stack that gets you hired in the AI era.
Twelve skills, four phases - from HTML to AI agents. For each one: what to learn, why it matters in 2026, and the exact tools you'll use. Tap what you've already got, watch your rank climb, then come build it for real.
Start unlockingThe real Sigmaschool curriculum, updated May 2026 for the agentic-AI era. Progress saves to your browser.
Deric Yee
Founder, Sigmaschool
Mentored hundreds of career changers into developer roles
This is the roadmap I wish someone handed me when I started.
It combines the best self-taught curriculum I have seen with the lessons I have learned mentoring hundreds of career changers into full-stack developer roles.
You will also find a practical sequence of projects, proof points, and technical milestones designed to help you build real skill - not just consume tutorials.
What this roadmap is for
A clear path for serious beginners who want to become AI-native software developers: people who can build, debug, deploy, explain, and use AI responsibly.
What this roadmap is not
This is not a vibe-coding shortcut. It is not about blindly asking AI to generate apps. It is about understanding software deeply enough to use AI as an amplifier.
Recommended resource map
As you work through the roadmap, use trusted references for the fundamentals. Start with official docs and battle-tested learning resources before chasing random tutorials.
- Deric
AI changes everything.
That is why fundamentals matter more.
AI is the elephant in the room, and pretending it is not changing everything would be dishonest. Yes, AI can generate code. Yes, it is getting better every month.
But that raises a different question. Not “Should you still learn to code?” The better question is “What kind of coding still matters?”
In 2026, the value is not in typing out syntax - AI can do that. The value is in understanding how software works, making decisions, debugging, shaping systems, and directing AI with clarity.
Shallow coding
People who do not understand the fundamentals become dependent on the tool. The value of typing syntax is collapsing.
Fundamental understanding
People who understand the fundamentals become dramatically more effective with AI. They work with it at a higher level of thinking.
The principle
Shallow coding is dying. Fundamental understanding is more valuable than ever.
What about vibe coding?
We are living in a generation of vibe coders: people who can spin up pages at lightning speed without understanding a single thing underneath. Tools are powerful. AI is powerful. Speed is useful. This roadmap is not about that.
This roadmap is about becoming a software developer - someone who understands how things work, why things break, and how to build a real career in technology.
Use AI as an amplifier, not a substitute for understanding.
AI as an amplifier
- Builds on mental models you already own
- Helps you debug, read docs, compare approaches
- Makes a real engineer dramatically faster
AI as a substitute
- Outsources the learning - no mental models form
- Ships AI slop: code with no thought or accountability
- Creates technical debt, not engineers
This is not anti-AI. You should use AI - every great engineer in 2026 uses AI in some form. The difference is simple: use AI as an amplifier, not a substitute for understanding.
No one is getting hired because they can produce a landing page with AI. That is not special, and it is not what the industry needs. The world still needs people who can think, reason, debug, design systems, and write clean, intentional code.
We have seen the consequences in interviews. Many graduates can generate snippets, copy patterns, or ask AI for answers, but cannot explain what their own code does. They cannot debug. They cannot reason through a problem.
Companies do not want button-clickers. They want thinkers - people who can build, diagnose, design, and make decisions. AI can help you go faster, but it cannot make you understand.
Why this roadmap exists
To rebuild the fundamentals. To train people who want to actually learn how to code - not vibe-code their way through it.
How long will this actually take?
At Sigmaschool, we tell students to generally plan for one year if they are working full-time and learning part-time. For part-time learners, that usually means 6–12 months depending on consistency, starting point, and weekly hours.
If you are learning full-time, the 90-day path is the intensive version - the goal is to compress the training by increasing focus, feedback, and weekly output.
This is the time required to build real skills the market values: fundamentals, debugging, systems thinking, deployment, communication, and AI-native workflows.
Full-time intensive
Working full-time + part-time learning
Casual learning
The real advantage
Embrace the fundamentals and you will outrun everyone later. Most people are trying to skip the boring parts - but the fundamentals are exactly what make you powerful with AI.
The Stack · Updated May 2026
What to learn, and why.
The AI-native developer stack.
Twelve modules across four phases - from HTML to AI agents. For each one: what you'll be able to do, why it matters in 2026, the exact tools, and roughly how long it takes. Tap each skill you've already got; progress saves to your browser.
Your rank
Curious explorer
Skills
0 / 12
Foundations
Build and ship your first real pages and apps.
HTML & CSS
⏱ ~30 hrs · ~5 daysStructure and style any interface from a wireframe.
WhyEvery screen you'll ever build is HTML + CSS underneath - you can't build or debug UI without it.
Lessons covered
- Semantic HTML structure & accessibility
- The box model - margin, padding, border
- Flexbox for one-dimensional layouts
- CSS Grid for two-dimensional layouts
- Responsive design & media queries
- Forms, inputs & their states
Fluent when you can
- Turn any wireframe into a pixel-accurate, responsive page
- Build mobile-first layouts that hold up on every screen size
- Debug a broken layout using browser DevTools
Free videos · Safa Yousif
JavaScript
⏱ ~50 hrs · ~8 daysRead, write, and debug real JavaScript - logic, arrays, objects, async.
WhyThe one language that runs the whole web. Master it once and everything after gets easier.
Lessons covered
- Variables, types & operators
- Functions, scope & closures
- Arrays & methods - map, filter, reduce
- Objects & manipulating the DOM
- Async - promises, async/await & fetch
- Error handling & debugging
Fluent when you can
- Read and explain unfamiliar JavaScript without guessing
- Manipulate the DOM and handle events to build interactivity
- Fetch and transform real API data with async/await
Free videos · Safa Yousif
Git & GitHub
⏱ ~15 hrs · ~3 daysBranch, commit, open PRs, and resolve conflicts like a team developer.
WhyEvery company runs on Git. No PR workflow, no job - it’s table stakes.
Lessons covered
- The command line & navigating a project
- init, stage, commit & reading history
- Branching & merging
- Pull requests & code review
- Resolving merge conflicts
- Remotes - push, pull, clone & fork
Fluent when you can
- Run a feature-branch → PR → merge workflow on a team
- Resolve a merge conflict without losing work
- Read a Git history to see what changed and why
Free videos · Safa Yousif
Real Apps
Ship production full-stack apps - with data and login.
React
⏱ ~40 hrs · ~7 daysBuild interactive UIs from reusable components with state and hooks.
WhyThe most-hired frontend skill in 2026 - it’s what product teams actually test you on.
Lessons covered
- JSX & your first components
- Props & composition
- State with useState
- Side effects with useEffect
- Lists, keys & conditional rendering
- Custom hooks & lifting state up
Fluent when you can
- Break any UI into a clean component tree
- Manage state and data flow without prop-drilling chaos
- Build interactive, data-driven interfaces from scratch
Free videos · Safa Yousif
Next.js
⏱ ~30 hrs · ~5 daysShip full-stack apps - routing, server rendering, and APIs in one framework.
WhyThe modern production standard. One framework takes you from a page to a deployed product.
Lessons covered
- App Router & file-based routing
- Layouts & nested routes
- Server vs client components
- Server actions & data mutations
- API routes
- Deploying to Vercel
Fluent when you can
- Ship a full-stack app from a single codebase
- Choose server vs client components correctly
- Deploy to production with a preview URL per branch
Free videos · Safa Yousif
APIs & data
⏱ ~20 hrs · ~3 daysRead API docs and wire any third-party service into your app.
WhyReal apps are glued together by APIs. If you can read docs, you can integrate anything.
Lessons covered
- HTTP, methods & status codes
- Reading API documentation
- fetch, headers & query params
- Working with JSON data
- Auth tokens & API keys
- Loading, error & empty states
Fluent when you can
- Integrate any documented third-party API into an app
- Handle every response state - loading, error, empty, success
- Keep secrets and API keys out of client code
Free videos · Safa Yousif
Databases & auth
⏱ ~30 hrs · ~5 daysModel data, run a real schema, and put it behind a login.
WhyEvery real product stores user data behind auth - this is the backend half of full-stack.
Lessons covered
- Relational data & schema design
- SQL - select, insert, update, delete
- Supabase setup & queries
- Authentication & sessions
- Row-level security & policies
- Connecting the frontend to the database
Fluent when you can
- Design a schema and run full CRUD against it
- Put data behind a login so users see only their own
- Reason about access control with row-level security
Free videos · Safa Yousif
AI Products
Build AI into products and direct coding agents.
AI pair-programming
⏱ ~15 hrs · ~3 daysShip 3× faster with Cursor + Claude - and know when to override them.
WhyIn 2026 the bar isn't 'can you code' - it's 'can you ship with AI and catch where it's wrong'.
Lessons covered
- Set up your Cursor + Claude workflow
- Prompting for code that actually works
- Generating, editing & refactoring with AI
- Reviewing AI output critically
- AI-assisted debugging
- Knowing when to override the AI
Fluent when you can
- Ship features noticeably faster with an AI pair
- Spot where AI-generated code is wrong or unsafe
- Decide when to take over and write it yourself
LLM APIs & AI features
⏱ ~25 hrs · ~4 daysIntegrate an LLM into a real app - prompts, streaming, evals, and cost.
WhyAn 'AI feature' is now a baseline product expectation. This is what makes you a 2026 hire.
Lessons covered
- Calling the Claude / OpenAI API
- Prompt & system-message design
- Streaming responses to the UI
- Structured output & tool use
- Evaluating quality & catching hallucinations
- Managing cost & latency
Fluent when you can
- Embed a working AI feature into a real product
- Design prompts that produce reliable output
- Evaluate AI quality instead of trusting it blindly
AI agents & agentic workflows
⏱ ~20 hrs · ~3 daysDirect a coding agent through spec → plan → diff-review → deploy.
WhyThe frontier skill: you set the spec and review the diffs while the agent does the typing.
Lessons covered
- The agent loop - think, act, observe
- Tool / function calling
- Writing a spec & plan for an agent
- Reviewing agent diffs
- Stop conditions, retries & guardrails
- Shipping agent-built changes
Fluent when you can
- Direct an agent through spec → plan → diff → deploy
- Review and reject bad diffs instead of rubber-stamping
- Build guardrails so an agent fails safely
Capstone
Ship a portfolio that gets you hired.
Ship & deploy your capstone
⏱ ~35 hrs · ~6 daysTake a vague brief all the way to a deployed, AI-native product.
WhyA portfolio of shipped products is your proof - it beats any certificate in an interview.
Lessons covered
- Scoping a vague brief into an MVP
- Building the core features
- Testing your app
- CI/CD & deployment
- Monitoring & rollbacks
- Writing the case study
Fluent when you can
- Take a vague idea to a deployed, working product
- Set up CI/CD so every push ships safely
- Recover from a bad deploy with a rollback
Free videos · Safa Yousif
Portfolio & job hunt
⏱ ~25 hrs · ~4 daysDefend your work on camera, prep interviews, and run the job hunt.
WhySkills don't matter if you can't defend them. This is exactly where solo learners stall.
Lessons covered
- Building your portfolio site
- Writing project case studies
- Recording demo walkthroughs
- Resume & LinkedIn for AI-native roles
- System design basics
- The interview process & mock interviews
Fluent when you can
- Defend your technical decisions on camera
- Present a portfolio that beats a certificate
- Run a structured job hunt and pass interviews
Free videos · Safa Yousif
Want the same path broken into 346 lessons and days you can tick off? Play it as a mission map.
Open the mission mapPhase lesson counts come from the real Sigmaschool curriculum. Per-module hours are focused-learning estimates; days assume the programme's ~30 hrs/week pace (~6 focused hours a day). Times vary with your starting point - treat them as a guide, not a guarantee. Stack and tools current as of May 2026.
What now
You've seen the path. Walk it in 12 weeks.
The roadmap is yours to keep. But the fastest way through it - every mission, sequenced and mentor-reviewed - is the programme.
AI-Native Software Developer Programme
From this roadmap to job-ready - for real.
- The same 12 missions - sequenced, mentor-reviewed, one a week
- AI-native from day one: Cursor, Claude, and agentic workflows
- Real portfolio + capstone, not certificates
- Job-ready in 12 weeks, with a money-back guarantee