Full catalog
Search 中文
More
AI VISTA / SITE NAVIGATION

Courses and topics open directly from the main navigation.

Open the full content catalog

Courses

4
  1. 01Build your first testable AI workflow5 lessons
  2. 02Make AI answers reliable enough to use14 lessons
  3. 03Design answers that can show their evidence11 lessons
  4. 04Automate with permission, recovery, and cost controls37 lessons
View all courses →

Topics

9
  1. 01Task framing4 guides
  2. 02Context and prompts6 guides
  3. 03Evaluation7 guides
  4. 04Grounded answers11 guides
  5. 05Agent architecture12 guides
  6. 06Agent safety5 guides
  7. 07Local model deployment7 guides
  8. 08Model operations7 guides
  9. 09Field cases4 guides
View all topics →

AI VISTA COURSES

Build an AI workflow step by step, like taking a real course.

Browse every course and its complete syllabus below. Compare who each course serves and what you will make, then enter at the right lesson.

4complete courses67sequenced lessons1real task to begin
Browse all courses ↓Open catalog →
LEARNING ARC / FOUR COURSESFROM TASK TO SAFE OPERATION
RECOMMENDED SEQUENCEEnter where your problem begins
01 5 lessonsBuild your first testable AI workflow 02 14 lessonsMake AI answers reliable enough to use 03 11 lessonsDesign answers that can show their evidence 04 37 lessonsAutomate with permission, recovery, and cost controls
EVERY COURSE LEAVES YOU WITH
01A syllabus
02Working artifacts
03Acceptance checks

ALL COURSES / COMPLETE SYLLABI

Choose a course, then open the lesson you need.

Every course shows its audience, outcome, and all published lessons.

60–90 minutes5 lessons

Build your first testable AI workflow

Move from an attractive idea to a bounded task, a workflow-or-agent decision, and a ten-case pilot a colleague can review before further investment.

Show all 5 lessons
  1. 01Should this task use AI at all?
  2. 02Turn an AI idea into a testable task brief
  3. 03Workflow or AI agent? Choose the control loop before building
  4. 04Draw the data boundary before AI sees the input
  5. 05Run a ten-case AI pilot before you scale
FOR
For anyone turning a repeated knowledge task into a first AI-assisted workflow.
OUTCOME
A task brief, control-loop decision, and ten-case pilot sheet.
LESSONS READ0 / 5
View syllabusStart course →
2–3 hours14 lessons

Make AI answers reliable enough to use

Preserve the right context, define content and format gates, test untrusted and difficult inputs, then diagnose failures in evidence, instructions, parsing, or handoff.

Show all 14 lessons
  1. 01Map the context before rewriting the prompt
  2. 02How many tokens should one LLM call reserve for input and output?
  3. 03When should you compact a long AI conversation?
  4. 04Write the acceptance contract before the prompt
  5. 05Choose JSON Schema, Markdown, or plain text for AI output
  6. 06Test prompt injection in documents and tool outputs
  7. 07Run a ten-case AI pilot before you scale
  8. 08Build an AI evaluation set that stays useful
  9. 09How do you calibrate an AI grader against human review?
  10. 10How to set AI classification thresholds with a human review queue
  11. 11How to test LLM temperature and sampling settings
  12. 12Diagnose an AI answer before changing the model
  13. 13The eval that would have caught this: turn one bad output into a regression test
  14. 14Monitor an AI workflow before drift becomes an incident
FOR
For teams whose prompt works in a demo but changes quality across real inputs.
OUTCOME
A context map, compaction checkpoint, acceptance contract, output protocol check, injection test set, and failure log.
LESSONS READ0 / 14
View syllabusStart course →
2–3 hours11 lessons

Design answers that can show their evidence

Choose an answer method, approve source versions, then test chunking, retrieval, access, index updates, citations, freshness, and no-answer behavior.

Show all 11 lessons
  1. 01Prompting, RAG, or fine-tuning? Choose by failure mode
  2. 02Decide whether RAG is worth building
  3. 03How do you approve sources before adding them to a RAG knowledge base?
  4. 04How do you validate PDF, table, and image extraction before RAG indexing?
  5. 05How should you chunk RAG documents? Keep the full evidence intact
  6. 06Keyword, vector, hybrid, or reranking for RAG? Choose with a matched trial
  7. 07Evaluate RAG retrieval before judging the answer
  8. 08Document-level access control for enterprise RAG, from search to citation
  9. 09How do you verify that a RAG document update or withdrawal reached the index?
  10. 10When RAG lacks evidence, should it answer, clarify, or hand off?
  11. 11Test citations for support and freshness
FOR
For knowledge assistants, policy search, support systems, and internal research workflows.
OUTCOME
A method decision, source admission register, chunking and retrieval trials, access and index checks, a no-evidence response gate, and a citation audit.
LESSONS READ0 / 11
View syllabusStart course →
Core 3–4 hours; branches vary37 lessons

Automate with permission, recovery, and cost controls

Build a bounded tool-using workflow with control ownership, data and action authority, stop rules, human review, recovery, service fallback, cost, and drift checks. Test delegation, MCP, memory, models, and deployment only when your task needs them.

Show all 37 lessons
  1. 01Workflow or AI agent? Choose the control loop before building
  2. 02Draw the data boundary before AI sees the input
  3. 03Give an agent the smallest useful permission
  4. 04How do you test an AI agent sandbox's file, network, and credential boundaries?
  5. 05Design an AI agent tool contract before exposing the function
  6. 06When should an AI agent continue, retry, or stop?
  7. 07Put human review where it can change the outcome
  8. 08Design recovery before the agent fails
  9. 09How should an AI model service fail over? Plan retries, queues, and human handoff
  10. 10Map the full cost of an AI workflow
  11. 11Your AI agent's real cost is cost per accepted task
  12. 12Monitor an AI workflow before drift becomes an incident
  13. 13How to implement an AI agent tool-calling loop
  14. 14Subagent or multi-agent? Choose the delegation topology
  15. 15When should an AI agent save, correct, or delete long-term memory?
  16. 16How can a long-running AI agent task resume without repeating work?
  17. 17How to debug one AI agent run from its observable trace
  18. 18What should you check before connecting a third-party MCP server?
  19. 19When should a repeated AI task become an agent skill?
  20. 20How do you load and unload an agent plugin safely?
  21. 21Your MCP server works locally. What breaks when a team depends on it?
  22. 22Claude Code without the chaos: a review-first workflow for coding agents
  23. 23Log the decision, not the prompt: a trace schema for AI agents
  24. 24Choose a model with a job-specific scorecard
  25. 25How much VRAM does a local LLM need? Budget weights and KV cache
  26. 26Is 4-bit LLM quantization worth it? Test fit, quality, and speed
  27. 27LoRA or QLoRA fine-tuning? Run a small, reviewable pilot
  28. 28Ollama, llama.cpp, vLLM, or SGLang? Choose a local inference engine
  29. 29How many concurrent requests can a local LLM serve?
  30. 30AI inference latency debugging: trace the wait from request to screen
  31. 31How many GPUs does a local LLM service need?
  32. 32Local LLM or cloud API? Choose by data, peak load, and operations
  33. 33Why prompt caching misses: run a three-request test
  34. 34AI support triage: the 10-case pilot that earned a review gate
  35. 35AI content review: the citation gate that stopped a confident wrong answer
  36. 36AI research synthesis: when the source changes after the brief
  37. 37The approval agent that could move money—until we removed the payment tool
FOR
For teams moving an agent or tool-using workflow from prototype toward production.
OUTCOME
A control-loop choice, data and permission boundaries, sandbox and tool contracts, runtime stop and human review gates, recovery runbook, service fallback, cost map, and drift watchboard; branch evidence where relevant.
SYLLABUS READ0 / 37
View syllabusChoose a path →

LEARNING ROUTE / STAGE 03 · COURSES

This stage follows a course path, then branches where the task requires it.

Courses are organized around real work. Related lessons reuse earlier decisions or artifacts. Complete the path your task needs and open relevant branches to produce a reviewable result.

BRING
The task direction found in the decision map
DO HERE
Complete decisions and outputs lesson by lesson
LEAVE WITH
A connected set of reviewable artifacts
WHEN THIS STAGE IS DONE Read for a specific gap Use the reading list when a branch needs more depth.

CHOOSE THE RIGHT COURSE

Choose by the problem in front of you—not by beginner or advanced labels.

The four courses address task fit, output reliability, evidence, and controlled automation. The automation course adds delegation and deployment branches for tasks that need them.

IF YOU ARE SEEINGSTART WITHDONE WHEN
01
You are turning a repeated knowledge task into a first AI workflow
Build your first testable AI workflow

Another person can run the pilot and reach the same pass/fail decision.

↗02
The prompt looks good in a demo but varies across real inputs
Make AI answers reliable enough to use

Each failed case is assigned to a fixable category instead of a vague quality complaint.

↗03
Answers must cite evidence that is traceable and current
Design answers that can show their evidence

A reviewer can trace every important claim to current, permitted evidence without exposing restricted source material.

↗04
An agent is about to use tools, write to systems, or change external state
Automate with permission, recovery, and cost controls

Every automated action has an owner, approval rule, observable result, and recovery path. A reviewer can validate the core controls and the branch evidence relevant to the task.

↗

HOW EACH COURSE WORKS

A lesson is not an article you finish. It is a small read, make, review, continue loop.

All four courses use the same learning actions on different problems. Bring one real task and advance one decision per lesson; complete a specialized branch only when its conditions apply.

  1. 01
    READ

    Identify the one decision this lesson changes

    Compare the case and boundary with real work; do not try to memorize every term.

    DONE SIGNAL · You can state the decision in one sentence.
    →
  2. 02
    MAKE

    Replace the blank tool with your own material

    Use current inputs, criteria, owners, and failure handling instead of copying the example.

    DONE SIGNAL · The artifact contains real fields, cases, or constraints.
    →
  3. 03
    REVIEW

    Ask a second person to apply the same standard

    Offer no verbal explanation. Let the reviewer decide pass, revise, or stop from the artifact alone.

    DONE SIGNAL · Disagreement becomes a named issue for the next iteration.
    →
  4. 04
    CONTINUE

    Carry the artifact into the next lesson

    Keep the same task. Open the reading list only when practice exposes a precise gap.

    DONE SIGNAL · The syllabus shows connected changes, not scattered notes.
    ✓