Free and open source
Free engineering courses that explain every step, quiz you while you learn, and bring the important ideas back before you forget them. You always know whether you understood the lesson or only recognised the words.
Every roadmap starts from its stated prerequisites and grows toward real work. Reviewed lectures are published continuously, while the complete planned path stays visible on every roadmap.
HTML to production full-stack
Start with a useful page, then build through JavaScript, TypeScript, React, APIs, databases, identity, Next.js, payments, real-time systems, testing, delivery, AWS, and GCP.
Prerequisites: None. The course begins with the browser, files, and a first published page.
Python to analysis and BI
Investigate real data from Day 1, then master Python, NumPy, Pandas, Excel, SQL, statistics, experiments, Tableau, metrics, and decision-ready analytical communication.
Prerequisites: None. Python, statistics, and analyst thinking are taught from first principles through real investigations.
First program to production systems
Compile a useful program on Day 1, then deepen through values, collections, domain design, concurrency, testing, HTTP, PostgreSQL, security, containers, and production operation.
Prerequisites: Basic programming fluency is recommended. The Python foundations in Python & Data Analytics are a suitable bridge.
From scratch to competitive
C++ from zero through the full data-structures and algorithms surface, including segment trees, advanced graphs and DP, number theory, and contest craft.
Prerequisites: None. C++ and problem-solving habits begin from zero.
Modern AI engineering
The maths that matters, neural networks built from scratch, then NLP, transformers, LLMs, RAG, agents, diffusion and the operational side of shipping models.
Prerequisites: Required: comfortable Python, NumPy, pandas, charts, descriptive statistics, and analytical reasoning. Complete Python & Data Analytics first if any of these are new.
React to shipped mobile apps
Build a real native screen, then deepen through adaptive layout, navigation, offline sync, device capabilities, accessibility, testing, native modules, and store delivery.
Prerequisites: Required: JavaScript and React fundamentals. Complete Web Development through the React phase first if these are new.
Missing a lesson or course? Request it and vote.
A subject can be organised neatly and still be taught badly. decko follows the order in which a learner can actually build the idea: a concrete problem, a useful mental model, a small example, a realistic example, the failure modes, and then practice.
Every lecture has one practical test: you should be able to read it from top to bottom without opening another tab to fill in a missing step. The rules that enforce that standard are public in the repository and checked before a lecture ships.
You begin with the version a reasonable beginner would write. It works. One realistic change breaks it. The new tool arrives only after you can feel why it is needed, so the rule has somewhere to attach.
A screen that makes three points is really three screens. Working memory holds very little at once, so material that arrives in dense blocks is mostly discarded before it is ever understood. Everything here is split until each screen carries exactly one idea.
Every example is buildable from what came before it. No API appears before it is taught, and nothing is waved away as something to understand later. If a term is needed early, it gets a plain-language detour first rather than a promise.
Rereading gets easier because the page becomes familiar. That does not prove the idea is yours. Lectures ask you to predict, trace, answer, and build before the explanation can make the work feel obvious.
The learning loop
Checkpoint and end-of-lecture quizzes are not decoration, and they are not saved for exam day. They are part of how each idea is taught and corrected while the reasoning is still fresh.
The lecture derives the idea from a problem you can see. Mental models come before syntax, and examples grow one careful step at a time.
Checkpoint quizzes appear between lecture sections. You answer before the explanation, while there is still time to repair the exact idea that did not land.
An end-of-lecture quiz mixes the concepts together. It tests whether the lesson still works when the nearby explanation is no longer doing the thinking for you.
Revision asks you to recall the idea after a gap. Missed material returns before the session ends, then comes back again after more time has passed.
Revision
A major review of common study techniques gave its strongest rating to two approaches: practice testing and distributed practice. Rereading, highlighting, and summarising received much weaker support, even though they are the habits most courses leave you with.
Revision here is not another page to consume. It asks before it explains, makes you answer from memory, and repeats what you missed before the session closes. The effort is useful because it reveals what survived and strengthens your ability to retrieve it again.
Dunlosky et al. (2013), Improving Students’ Learning With Effective Learning Techniques; Roediger & Karpicke (2006), Test-Enhanced Learning. The full reasoning is in the repo.
Thirty seconds of context so you know what is about to be asked. It gives you a direction without giving away the answer.
Cards ask first and explain second. You grade yourself honestly, and anything you miss returns before the session closes.
A graded quiz unlocks after recall is complete. You cannot skip straight to recognition and mistake it for memory.
History and scores show how long it has been. Returning after a gap gives the next retrieval attempt something meaningful to do.
Nothing is generated while you wait. Every lecture and revision set is reviewed and committed before a learner sees it. You get stable material that can be corrected for everyone, not a fresh draft generated while you wait.
One row per habit. Mark the day, keep the run alive.
Loading…
decko ran privately for about eight months before it was opened. Early lectures were too shallow. Some examples depended on ideas from later phases. Some paragraphs explained what code did and never earned the learner’s understanding of why.
Those failures became rules. Phases were reordered, lectures were rebuilt, quizzes moved inside the teaching flow, and revision was redesigned around retrieval. The result is not just a large content library. It is a public system for producing explanations that can be inspected and improved.
decko is free and open source. The application, curriculum, teaching rules, and every lecture are available to read, fork, correct, and build on.