A complete, self-contained study site for Anthropic's Claude Certified Architect – Foundations (CCA-F) exam — built by a complete beginner (with an AI tutor) and designed to be dyslexia-friendly: diagram-first, plain words, big readable text.
⚠️ Community project. Not affiliated with or endorsed by Anthropic. Exam details shown in the course are community-reported — always verify with Anthropic (academy-support@anthropic.com) before booking anything.
- ⚛️ The Atomic Curriculum — 23 prerequisite-aware units, starting with a five-lesson computer-skills Workbench and building through Claude foundations, agentic systems, reliability, and production scenarios
- 🛤️ Timeline — the whole course as one visual path with ✓ done / ⭐ you-are-here tracking
▶️ Lesson page — tells you exactly what to do next: your current lesson, its video, and its 5 steps (they stay with the lesson until you finish it)- 🔊 Natural read-along — the seven beginner Workbench lessons include local ElevenLabs narration, word highlighting, a seek bar, and remembered
0.75x-2xspeed control - 🎥 Reviewed video library — 22 recent community episodes catalogued, with 14 exact lesson clips, full-episode links, transcript provenance, and current-product warnings
- 🗺️ Prerequisite concept map + 3 engineering articles retold as ~30 diagrams (Building Effective Agents, Multi-Agent Research System, Writing Tools for Agents — original links included)
- 🃏 5 flashcard decks with two modes: flip-through and a Quizlet-style Learn mode that re-asks what you miss
- 🧪 Pre-test (30 Q) + 6 checkpoint quizzes that track your score and how much you guessed
- 🏋️ One fixed build per unit with worked examples, independent redo, progressive hints, verification steps, and a four-part rubric
- 🗣️ Typed or spoken teach-back + a simple spaced-review queue for concept and mistake cards
- 💬 Grounded tutor bridge that copies the current lesson, prerequisite, rubric, and recent evidence into a focused tutor prompt
- 🧰 Applied Engineer path for programming, APIs, databases, Git, testing, systems, security, architecture, and technical-cofounder judgment
- 🩹 Repair guide — the trickiest pre-test questions explained as pictures
- 🎛️ NEW: CCA-F Study Studio — a rebuilt guided app (React + a local Python server + a private SQLite database) that walks one unit at a time through Learn → Draw → Build → Teach → Quiz → Review, with server-side grading, verified backups, and an MCP connection for AI tutors (FRONTIER-AI.md). Unit W1 is fully interactive today; the other 22 units are in the protected manifest for staged migration, and the classic pages above stay available at
/legacy/. - Progress is saved automatically when you run the course locally — every tick, card and typed answer; no account, no tracking. The classic pages save to your browser and (with
serve.py) to a real file,my-progress.json, which auto-restores if the browser's copy is ever cleared; Study Studio saves to a private local database with verified export/import backups. The hosted site is a preview only and does not save progress.
Open https://zengdaniel7.github.io/claude-architect-study-course/ — a preview of the new Study Studio. You can click through the first unit (W1) end to end: Learn → Draw → Build → Teach → Quiz → Review. It shows "Preview only. Progress is not saved." and means it — nothing persists between visits, and the AI tutor is switched off. Use it to see the course; download it to study.
git clone https://github.com/zengdaniel7/claude-architect-study-course.git
cd claude-architect-study-course
python3 serve.pyOpen http://localhost:8000 — start at the Home page. Downloading unlocks what the hosted preview can't do:
- Your progress becomes a real file —
serve.pymirrors every tick and typed answer intomy-progress.jsonin the folder, automatically, and restores from it if browser data is ever cleared. (Plainpython3 -m http.server 8000works too, but then saves stay browser-only.) - The course becomes YOURS to grow — run an AI coding agent (like Claude Code) in the folder and the included CLAUDE.md turns it into your personal tutor: it grades your exercise reports, keeps your mistake log, adds new flashcards and exercises to the actual files as you learn. The hosted site can never edit itself — a local copy can.
- Works offline, and you can edit any lesson to match how you learn.
⭐ Want the new Study Studio app locally? It ships as source. Build it once with Node 20+ and pnpm (corepack enable && pnpm install && pnpm run studio:build), then launch:
./Start\ CCA-F\ Study\ Studio.commandOpen http://127.0.0.1:8765/ — W1 fully guided with server-side grading, progress in a private local database (with verified export/import backups), the classic pages at /legacy/, and MCP hookup for AI tutors (FRONTIER-AI.md).
| 🌐 Hosted preview | 💻 Downloaded | |
|---|---|---|
| See the new Study Studio (W1 guided journey) | ✅ preview | ✅ (after one build) |
| Full classic course: lessons, quizzes, flashcards, exercises | ❌ | ✅ |
| Progress saves | ❌ never (preview only) | ✅ browser + my-progress.json, or the Studio database |
| AI tutor that grades + upgrades the course files | ❌ | ✅ |
| Edit lessons / add your own content | ❌ | ✅ |
| Works offline | ❌ | ✅ |
The course is model-agnostic. Every "Ask my tutor" and "🧠 Check my answer" button copies a ready-made prompt to your clipboard — paste it into any AI you use (Claude, ChatGPT, an AI browser like Comet, or a local model). The prompt includes everything the AI needs, including hidden grading rubrics for exercises.
Power mode: run an agentic coding CLI (like Claude Code) inside this folder. The included CLAUDE.md briefing turns it into a full tutor that can grade your exercise reports, maintain a mistake log, add flashcards, and grow the course as you learn. It works with other agent CLIs too (point them at CLAUDE.md).
- Open Home → take the Pre-Test to set your baseline (watch your guess count, not just the score)
- Every day, open Today — watch the short assigned clip first, open the full episode when useful, then continue through notes → draw → build → explain → flashcards
- Complete the unit's independent build and 60-second teach-back
- Retake its quiz until you reach at least 80% with zero guessing
- A unit is mastered only when all four evidence types are present: lesson steps, quiz, independent build, and teach-back
- Curriculum structure inspired by Atomic Design by Brad Frost
- Accessibility baseline documented in DESIGN-STANDARD.md
- Video choices and timestamps are documented in VIDEO-CURRICULUM-MAP.md; community videos are clearly separated from official Anthropic sources
- Diagrams retell three Anthropic engineering articles (linked at the top of each article page)
- Font: Atkinson Hyperlegible (Braille Institute, SIL OFL) — designed for low-vision and dyslexic readers
- Foundation narration: generated through Higgsfield using an ElevenLabs preset voice; audio is stored locally and no API key is shipped in the site
- Free prep courses: Anthropic Academy
Code and course text: MIT. The Atkinson Hyperlegible fonts in fonts/ are licensed separately under the SIL Open Font License (not MIT). "Claude" and "Anthropic" are trademarks of Anthropic, PBC — this project is unaffiliated.