About Courses Programs Contact Log in Sign up

Home›Advanced Prompt Engineering›29 - Lost in the middle

Loading the video…

🔒

Log in to watch this video

All Académie WS videos are free, but reserved for members. Signing up takes less than a minute.

Log in Create an account

29 - Lost in the middle

Artificial IntelligenceLessonFree

Next chapter →☰ Syllabus ⬇ Course material ← Previous

Chapters

68 lessons

01 - The promise of the course: prompting, context, models, cost and responsible use

02 - Who I am and where to follow me

03 - What you get at the end: the framework, the checklists and a reusable prompt

04 - The roadmap of the 8 modules

05 - Good vs bad prompt: why prompt B wins

06 - Module intro: understanding what's behind ChatGPT, Claude and Copilot

07 - What an LLM is, and the difference between tool and model

08 - Temperature: precise or creative

09 - Module intro: going from a vague request to a professional prompt

10 - 4 terms: prompt, prompt engineering, specificity, instruction hierarchy

11 - The anatomy of R-C-T-C-O

12 - From a weak prompt to a structured prompt

13 - Example: a customer email built step by step

14 - RTF, the minimum for everyday requests

15 - CRAFT and RISEN for richer tasks

16 - Decomposition and iterative improvement

17 - Module intro: advanced techniques and when to use them

18 - Zero-shot vs few-shot

19 - Negative examples: showing what to avoid

20 - Chain of Thought: reasoning step by step

21 - Personas: giving the AI a point of view

22 - Prompt chaining: one source, four outputs

23 - Module intro: a conversation is not infinite

24 - Tokens and the context window

25 - The limits of a conversation, ChatGPT vs Claude

26 - Memory: conversation, project or account

27 - Sorting context: essential, optional or noise

28 - The inverted pyramid

29 - Lost in the middle

30 - Delimiters

31 - Long documents: summarize, select, split, spread out

32 - Chunk & summarize vs targeted extraction

33 - Example: lightening an overloaded prompt

34 - Module intro: the right tool, the right model, at the right cost

35 - ChatGPT, Claude or Copilot: the differences

36 - The 4 criteria for choosing a model

37 - Which tool for your profile

38 - The multi-model decision framework

39 - Input vs output tokens: which costs more

40 - Managing costs: caching, model choice, output control

41 - Module intro: deciding the result before the AI writes

42 - The 6 levers to control the response

43 - Behavior constraints vs format constraints

44 - One single instruction for a fully controlled result

45 - Module intro: when to verify, and which data to entrust to which account

46 - Hallucinations and their signs

47 - Making the model show its evidence

48 - The 4 safeguards at work

49 - Protecting your data: what "training" means

50 - Each tool's guarantees depending on the account

51 - Individual plan vs enterprise plan

52 - Case: spotting a plausible but unsourced answer

53 - Six real cases: trust or verify?

54 - Module intro: tying everything into a single method

55 - The complete 6-step workflow

56 - Example: from rough notes to an executive brief

57 - Your reusable prompt template

58 - The 4 checklists

59 - Glossary (1/2)

60 - Glossary (2/2)

61 - Bonus intro: going further with your tool

62 - Setting up and using ChatGPT, Claude and Copilot well (1/6)

63 - Setting up and using ChatGPT, Claude and Copilot well (2/6)

64 - Setting up and using ChatGPT, Claude and Copilot well (3/6)

65 - Setting up and using ChatGPT, Claude and Copilot well (4/6)

66 - Setting up and using ChatGPT, Claude and Copilot well (5/6)

67 - Setting up and using ChatGPT, Claude and Copilot well (6/6)

68 - 7-day action plan and questions in the comments