Reliable AI-Assisted Software Engineering
Frontend, backend ও mobile developer সবাই নিজের stack-এ practice করতে পারবে। আগে basic coding ও Git জানা দরকার। Junior হলে ছোট exercise দিয়ে শুরু করো; senior হলে সিদ্ধান্ত, ঝুঁকি ও verification-এর প্রমাণ দেখাও। শুধু quiz পাস করা মানেই দক্ষতা অর্জন নয়।
What you will practise
Twelve lessons cover task definition, context engineering, tools and permissions, specifications, orchestration, bounded harnesses, evaluation, cost measurement, code review, production operations, legacy migration and career evidence. Each lesson offers frontend, backend and mobile examples. Complete one stack, not all three.
Before you start
- Be able to edit and run a small program, read an error and make a Git commit.
- Use a browser for the lessons. The optional downloadable lab needs Python 3 and a terminal. Its code is shared infrastructure practice, not a frontend or mobile app.
- Core fixture exercises need no API key or paid model. A live model adapter is an optional extension with provider costs and terms.
- Lessons combine English technical text with Bangla explanations; this is not a fully translated video course.
Your learning path
- Work through lessons 1–3: explain agent, workflow and harness; write a task brief and permission boundary.
- Lessons 4–6: specify one feature, divide work by contracts and run a bounded fixture loop.
- Lessons 7–9: prepare expected outcomes before testing, measure available usage and review negative/security cases.
- Lessons 10–12: rehearse release and rollback in a sandbox, preserve legacy behaviour and present a defensible case study.
Junior: complete each foundation exercise, explain one failure and its fix. Senior: add the extension and defend integration, security, cost and recovery decisions. Start with foundations whenever a concept is new.
One capstone, three choices
Build one slice of a synthetic support-ticket assistant. Use made-up data. Any outgoing reply requires human approval.
- Frontend: an accessible ticket form and draft-review screen against fixture APIs. Demonstrate loading, error and keyboard-only flows.
- Backend: ticket storage and classification behind authentication and authorisation. Demonstrate duplicate handling and denial of another user's records.
- Mobile: offline ticket capture and sync against fixture APIs. Demonstrate restart, retries and duplicate prevention.
Download the Python offline starter. Extract it, read README.md, then run python3 assistant_lab.py and python3 -m unittest -v inside that directory. It uses trusted session fixtures and in-memory storage. It is not a production API or a real model integration. Build your chosen UI/API separately.
What counts as completion?
Keep a small evidence folder with your task brief, acceptance cases, source changes, failing and passing test results, review notes and a five-minute explanation. Use this rubric with a colleague; it is a learning checklist, not an accredited assessment.
- Behaviour: show the expected outcome and at least one failure before its fix.
- Access: demonstrate a denied unauthorised operation.
- Cost: report actual usage when available; label fixture-only results and unavailable figures honestly.
- Operations: record an observable failure and demonstrate rollback in your sandbox.
- Judgment: explain a design tradeoff without AI writing the answer for you.
Quiz scores, XP or a platform completion badge do not establish professional competence. Do not put unfinished exercises or invented metrics on a CV.
Scope and known limits
This is a practical foundation and exercise track, not exhaustive software engineering training. It does not replace language/framework fundamentals, deep systems/security study or supervised production experience. It is not an ML model-training programme, a complete video bootcamp, or a guarantee of employment or immunity to automation. Stack examples are exercise prompts, not three fully implemented reference applications. Quizzes are basic knowledge checks; practical review is still needed. Independent learner validation has not yet been completed.
The material is original, informed by a public topic outline and linked official references. It is not affiliated with or endorsed by Philipp Lackner and does not reproduce his paid lessons or coaching.
Share and improve
Share this as a free practical self-study course for frontend, backend and mobile developers, with junior foundations and senior extensions. Try one lesson with a colleague and compare explanations and evidence.
Report a correction or confusing step. Include the lesson title, expected behaviour and a minimal synthetic example. Do not include secrets, customer records or private employer code. Keep your own copy of important learning notes.