Changelog

Anthropic revises exam guides without notice, and study material goes stale quietly. This page records what changed and when — including corrections to material that was wrong here first. If something you studied has since been fixed, you'll find it below rather than nowhere.

1. Site 29 October 2026

   CCAR-P (Architect Professional) track is live: all four certifications covered

   The fourth and final track. Lessons cover all 38 objectives in the CCAR-P exam guide v1.0, across solution design, models and context engineering, integration and RAG, evaluation, governance and compliance, stakeholder lifecycle, and team enablement, plus a practice bank, a 14-question diagnostic, a 63-item mock exam, and a quick reference. Every practice question has its own page.

   Integration is the heaviest domain at 19%. Governance and stakeholder communication are 28% combined.

   Source: Anthropic, Claude Certified Architect – Professional Exam Guide v1.0

2. Site 28 October 2026

   Every practice question now has its own page

   All 357 practice questions across CCDV-F, CCAR-F (including the six exam scenarios), and CCAO-F now live at their own URL, with the answer and the full explanation of why each wrong option is wrong. Link a friend straight to a question you found tricky, or work through a domain one question at a time. Each track’s practice page lists every question, grouped by domain.

3. Site 27 October 2026

   CCAO-F (Associate Foundations) track is live

   The third full track: a plain-language curriculum for people who use Claude at work rather than build with it, plus a practice bank, a 14-question diagnostic, a 60-item mock exam, and a quick reference.

   All seven domain names, objectives, and weights come from the CCAO-F exam guide v1.0 (effective July 2026). Output Evaluation and Validation is the heaviest domain at 21%. Prompting is 14%.

   Source: Anthropic, Claude Certified Associate – Foundations Exam Guide v1.0

4. Content 27 October 2026

   CCAR-F curriculum rebuilt: one lesson per exam-guide task statement

   The CCAR-F curriculum now has 30 lessons, numbered exactly like the guide’s 30 task statements, so lesson 3.4 is task statement 3.4 (plan mode vs direct execution). Before, our numbering reused the guide’s numbers for different topics, and 14 task statements had no lesson at all.

   New lessons: tool distribution and tool_choice (2.3), built-in tool selection (2.5), path-specific rules (3.3), plan mode (3.4), iterative refinement (3.5), Claude Code in CI/CD (3.6), explicit criteria (4.1), batch processing (4.5), multi-pass review (4.6), escalation (5.2), large-codebase context (5.4), human review and calibration (5.5), and provenance (5.6). Session resumption and forking (1.7) was rewritten.

   Retired, because the guide lists them as out of scope: MCP server hosting and transport, OAuth and API-key handling, and prompt-caching implementation and pricing. Old lesson URLs redirect to the closest new lesson.

   The practice bank is now 105 questions, at least three per task statement, with off-guide and out-of-scope questions removed.

   Source: Anthropic, Claude Certified Architect – Foundations Exam Guide v1.0, sections 6 and 17

5. Correction 27 October 2026

   Corrected: six CCAR-F lessons and three questions realigned with the exam guide

   An audit against the guide’s task statements found places where we taught the opposite of what the exam rewards. Fixed:

   • 1.1 Agentic loops. The loop now branches on stop_reason: continue on "tool_use", stop on "end_turn". An iteration cap is a safety backstop, not the primary stop. Questions arch-1-01 and arch-1-02 rewritten to match.
   • 1.3 Subagent context. Subagents start with isolated context; the coordinator must pass what they need explicitly. We had implied they inherit it by default.
   • 2.1 Tool descriptions. Differentiating tool names and descriptions is the right first fix for misrouting; merging tools is a heavier option. Question arch-2-01 rewritten. Extraction fields the source may lack should be optional or nullable, so the model doesn’t invent values.
   • 3.2 Skills. A skill runs in the main conversation by default; isolation (context: fork) and tool limits (allowed-tools) are opt-in. The trap was inverted.
   • 4.1 System prompts. Escalation criteria belong in the system prompt as explicit rules with examples, not sentiment (“customer is angry”). Rules that must never be broken, like refund eligibility, are enforced in code or hooks, not tool descriptions.

   New: scenario practice for all six exam scenarios, 60 questions.

   Source: Anthropic, Claude Certified Architect – Foundations Exam Guide v1.0, section 6

6. Correction 27 October 2026

   Corrected: the four exams don't share one fee

   The homepage and “Which Claude certification should you take?” said every exam costs US$125. That’s true for CCDV-F and CCAR-F only. Per each exam guide’s “Exam details at a glance” table, CCAO-F is US$99 and CCAR-P is US$175. Partner-tier discounts are applied at checkout.

   Source: Anthropic exam guides v1.0 (CCAO-F, CCAR-P)

7. Correction 27 October 2026

   Corrected: CCAR-F weights for Prompt Engineering and Context Management

   At launch we estimated Prompt Engineering & Structured Output at 19% and Context Management & Reliability at 16%, because we didn’t have those two figures. The exam guide v1.0 gives 20% and 15%. Every CCAR-F page, the diagnostic ranking, and the mock exam’s domain sampling now use the published weights. The other three domains (27%, 20%, 18%) were already correct.

   Source: Anthropic, Claude Certified Architect – Foundations Exam Guide v1.0

8. Content 19 September 2026

   CCAR-F (Architect, Foundations) curriculum launched — 27 lessons, 70 questions

   The second certification track is live: Claude Certified Architect (Foundations), free and open at /ccar-f, with its own curriculum, diagnostic, practice bank, and mock exam.

   27 lessons across 5 domains — Agentic Architecture & Orchestration, Claude Code Configuration & Workflows, Prompt Engineering & Structured Output, Tool Design & MCP Integration, and Context Management & Reliability — plus a 70-question practice bank.

   On the domain weights: Agentic Architecture & Orchestration (27%), Claude Code Configuration & Workflows (20%), and Tool Design & MCP Integration (18%) are drawn from the published blueprint. Prompt Engineering & Structured Output (19%) and Context Management & Reliability (16%) are our best estimate of the remaining share — we don’t have those two figures confirmed against the current exam guide yet. If you have the official CCAR-F exam guide and these are off, that’s exactly the kind of thing this changelog exists to correct.

9. Correction 19 September 2026

   Corrected: MCP remote transport is Streamable HTTP, not sockets

   Lesson 8.2 and one practice question described the remote MCP transport as “network sockets.” That was wrong. The specification defines two standard transports: stdio for a locally launched subprocess, and Streamable HTTP for remote and multi-client access.

   Also worth knowing, since older material still repeats it: HTTP+SSE is deprecated, replaced by Streamable HTTP as of spec version 2025-03-26. If an answer option offers SSE as the current remote transport, that’s the dated option rather than the correct one. The practice question has been rewritten so SSE is the distractor.

   Source: MCP specification 2025-03-26

10. Correction 19 September 2026

    Corrected: hook examples now use the real settings.json format

    Lessons 1.2 and 7.3 illustrated hooks with an invented Python decorator (@hook("pre_tool_use")). Hooks are configured as JSON in settings.json, events are PascalCase — PreToolUse, PostToolUse, UserPromptSubmit, Stop, SessionStart, PreCompact — and a matcher targets them at specific tool names.

    The handler receives the event as JSON on stdin and refuses either by exiting with code 2 or by printing hookSpecificOutput.permissionDecision. Exit 2 is the hard block: on events that can block it stops the call whether or not you print JSON.

    Both lessons now show the actual shape.

    Source: Claude Code hooks documentation

11. Content 19 September 2026

    Batch and caching figures replaced with verified numbers

    Several lessons said “a steep discount” where real, testable figures exist. Now stated:

    • Batch: 50% off both input and output tokens, applied automatically. 24-hour ceiling, not an estimate. Up to 100,000 requests or 256 MB. Results kept 29 days and returned out of order, matched by custom_id. No streaming inside a batch.
    • Caching: reads at 0.10× base input, writes at 1.25× for the 5-minute TTL or 2.0× for the 1-hour TTL. Reads reset the TTL.
    • The two stack — but the default 5-minute TTL expires partway through anything but a small batch, so use the 1-hour TTL inside one.

    The consequence worth carrying: a misplaced checkpoint is worse than no caching, because you pay the write premium on every call for a cache nothing ever reads.

    Source: Anthropic batch, pricing and prompt-caching documentation

12. Site 16 September 2026

    Diagnostic added — ranks domains by cost, not accuracy

    Twelve questions, no feedback until the end. The report ranks domains by weight × (1 − accuracy), so scoring 0% on Eval and Debugging ranks below scoring 50% on Agents and Workflows. Bombing a 2.6% domain costs you almost nothing; being shaky on a 33.1% domain costs you the exam.

13. Site 12 September 2026

    Curriculum live — 25 lessons, one per sub-skill

    Every sub-skill in the blueprint now has its own lesson with objectives, worked examples, a practice scenario, a build exercise and its exam traps. Lesson length is sized by domain weight rather than split evenly.

14. Blueprint 1 July 2026

    CCDV-F Exam Guide v1.0 published, effective July 2026

    First published blueprint for Claude Certified Developer – Foundations. Eight domains, 25 individually weighted sub-skills, 53 questions, 120 minutes, 720/1000 to pass.

    The weighting is unusually skewed: Applications and Integration alone is 33.1%, the heaviest single domain on any of the four Claude certifications. Claude Code and Eval/Testing/Debugging together come to 5.7%.

    This is the version all material on this site is written against.

    Source: Anthropic exam guide v1.0