The overview session of the no-code-first AI Foundations track, in 22 slides, holding the whole track in a single deck. It sets out the concepts-then-map-then-build philosophy and the "garbage in, garbage out" north star, then surveys what AI can build for you: chatbots, text processors, and automations. You will see the four ways to use Claude - a one-off chat, a saved Artifact, a reusable Skill, and the API - and learn to choose between them based on how often a tool runs and where it runs. The session closes with a live walk through the mapping worksheet (Goal, Inputs, Outputs, Constraints, Success test, Reuse) applied to one running need: turning messy meeting notes into a clean action-item list. It flags two traps, building before mapping and using a one-off chat for a repeated job, adds two checks and a reusable pattern for taking a fuzzy idea to a finished tool, and ends with a worksheet the student fills in for their own idea. Nothing needs installing and no code is required; everything runs in the Claude app on the $20 plan.
Subject: AI Foundations · 44 slides · applied lesson
Open the interactive version of this deck · Homework for this lesson
Title
Overview Session · No-code-first track
Before we build anything: know what's possible, map it out, then build it - so the inputs are good before the tool exists.
Objectives
This first session is the map of the whole track. Nothing to install - we work entirely in the Claude app you already have. By the end you can:
Concept
AI gives back the quality of what you put in. A vague ask produces a confident-sounding but useless answer. A clear, well-specified ask produces something you can actually use.
| What you give it | What comes back |
|---|---|
| "Help me with my notes." | A generic essay about note-taking - not your notes. |
| "Turn THESE messy meeting notes into a dated list of action items, one owner each." | Exactly that list, in the shape you asked for. |
The whole track is built to keep you on the right-hand column. We get the inputs good before anything is built.
Comparison
Comparison matrix
From Your north star: no garbage in, garbage out: refill the What comes back column from what you know. The rest of the table is as it appeared.
| What you give it | What comes back |
|---|---|
| "Help me with my notes." | A generic essay about note-taking - not your notes. |
| "Turn THESE messy meeting notes into a dated list of action items, one owner each." | Exactly that list, in the shape you asked for. |
Concept
We never jump straight to building. Everything we do this track is one of three moves, in order:
Today is mostly moves 1 and 2. Building comes once the map is solid - that's the point.
Counterexample
Discussion prompt
We never jump straight to building. Everything we do this track is one of three moves, in order:
That is stated as though it always holds. Do one of two things: produce a case where it fails, or say precisely what rules such a case out. "It just does" is not on the menu.
Hint: Hunt at the extremes first — zero, one, negative, empty, equal. If every extreme survives, the reason they survive is the proof.
Answer:
Today is mostly moves 1 and 2. Building comes once the map is solid - that's the point.
Section
The landscape
Concept
Strip away the hype and almost everything you'd want falls into three shapes:
Our running example - messy meeting notes into a clean action-item list - is a text processor. We'll carry it through the whole session.
Matching
Match the pairs
From What AI can build for you — match each one to what it actually does. The descriptions have been shuffled.
Why: Chatbots / assistants, Text processors, Automations are easy to tell apart while they are sitting next to their descriptions and much harder afterwards, which is what this checks.
Picture it
Figure (svg): Messy notes go into a box labeled the tool and a clean action list comes out the other side
Discussion prompt
Read the picture before the words. What is this showing, and what is the one thing it is built to make obvious? Commit to an answer, then read on.
Hint: Name the parts, then say what changes between them — and if nothing changes, say what is being held still.
Answer:
Don't let the names intimidate you. Under the hood each one is the same picture: something goes in, the tool does its job, something comes back out.
Intuition
Don't let the names intimidate you. Under the hood each one is the same picture: something goes in, the tool does its job, something comes back out.
Figure (svg): Messy notes go into a box labeled the tool and a clean action list comes out the other side
If you can describe the in and the out precisely, you can build the tool. That's the entire game.
Analogy
Discussion prompt
Explain Every one of them is just in -> out by analogy to something with no AI Foundations in it at all — a queue, a recipe, a map, a bank balance, whatever fits. Then say where your analogy breaks.
Hint: An analogy that never breaks is not an analogy, it is the same idea wearing a hat. Find the seam — that is the part that is actually new.
Answer:
Don't let the names intimidate you. Under the hood each one is the same picture: something goes in, the tool does its job, something comes back out.
Section
Four ways to use Claude
Concept
Same AI underneath - four different ways to reach it. Three live right in the Claude app on your $20 plan; the fourth is for later, and only if a project needs it.
| Approach | What it is | Best for |
|---|---|---|
| Chat | A normal conversation in the Claude app. | Thinking, drafting, a one-off question. |
| Artifact | A thing Claude builds in the chat you can keep and reopen. | A document, template, or small app you'll return to. |
| Skill | Saved instructions that teach Claude your repeated task. | A job you do over and over, done the same way each time. |
| API | Wiring Claude into other software / automations. | Runs automatically, at scale, or inside another app. |
The API is the only one that crosses into developer territory - that's why we don't touch it until Session 4, and only if your project actually needs it.
Comparison
Comparison matrix
From The four ways to use Claude: refill the Best for column from what you know. The rest of the table is as it appeared.
| Approach | What it is | Best for |
|---|---|---|
| Chat | A normal conversation in the Claude app. | Thinking, drafting, a one-off question. |
| Artifact | A thing Claude builds in the chat you can keep and reopen. | A document, template, or small app you'll return to. |
| Skill | Saved instructions that teach Claude your repeated task. | A job you do over and over, done the same way each time. |
| API | Wiring Claude into other software / automations. | Runs automatically, at scale, or inside another app. |
Picture it
Figure (svg): A decision branch: run it once leads to chat or artifact, run it often your way leads to skill, runs automatically inside other software leads to API
Discussion prompt
Read the picture before the words. What is this showing, and what is the one thing it is built to make obvious? Commit to an answer, then read on.
Hint: Name the parts, then say what changes between them — and if nothing changes, say what is being held still.
Answer:
You don't memorize a chart. You ask two questions: How often will I run this? and Where does it need to run?
Intuition
You don't memorize a chart. You ask two questions: How often will I run this? and Where does it need to run?
Figure (svg): A decision branch: run it once leads to chat or artifact, run it often your way leads to skill, runs automatically inside other software leads to API
This is the same "Reuse" question from the worksheet you'll meet in a minute. It quietly decides almost everything.
Analogy
Discussion prompt
Explain Two questions decide it by analogy to something with no AI Foundations in it at all — a queue, a recipe, a map, a bank balance, whatever fits. Then say where your analogy breaks.
Hint: An analogy that never breaks is not an analogy, it is the same idea wearing a hat. Find the seam — that is the part that is actually new.
Answer:
You don't memorize a chart. You ask two questions: How often will I run this? and Where does it need to run?
Ranking
Put in order
Put the moves of Pick the tool for our notes into the order they have to happen.
Why: These are the moves of the worked example in the order it makes them, and each one is set up by the one before it. Repeated use rules out a throwaway chat I'd have to re-explain every time.
Worked example
Apply the two questions to messy meeting notes -> action-item list. Decide before you read the next beat.
Ask: how often? I run a meeting most weeks, so - repeatedly.
Why: Repeated use rules out a throwaway chat I'd have to re-explain every time.
Ask: where? Right in the Claude app, by hand, when the meeting ends.
Why: No other software has to call it automatically - so I don't need the API.
Land on a Skill (an Artifact template is the lighter first step).
Why: Repeated + same shape + lives in the app = a Skill. We'll likely prototype the format as an Artifact first, then save it as a Skill.
Reverse engineer
Discussion prompt
Work backwards. The example finished here:
Land on a Skill (an Artifact template is the lighter first step).
What was it asked to do, and what must it have been given? Reconstruct the problem from its answer.
Hint: Every quantity in the result had to enter somewhere. Account for each one.
Answer:
Apply the two questions to messy meeting notes -> action-item list. Decide before you read the next beat.
Anomaly
Predict first
A student writes this, and it looks reasonable:
Treat the weekly notes job as a fresh chat every time.
It is wrong. Say what breaks — and say it before you turn the page.
Correct: Nothing is saved. You retype the instructions, and small wording changes make the output drift week to week.
Save the format once so it runs the same way every time.
Why: Nothing is saved. You retype the instructions, and small wording changes make the output drift week to week.
Trap
Treat the weekly notes job as a fresh chat every time.
Each week: open a blank chat and re-explain the format you want
Why: Nothing is saved. You retype the instructions, and small wording changes make the output drift week to week.
Result: inconsistent lists and wasted time
Why: You're rebuilding the same tool from scratch every Monday.
Save the format once so it runs the same way every time.
Capture the instructions as a Skill (or a reusable Artifact template)
Why: The format lives in one place - you invoke it instead of re-describing it.
Result: same clean shape every week, in seconds
Why: Repeated work belongs in something reusable, not a disposable chat.
Break the constraint
Discussion prompt
The rule this trap just fixed:
The format lives in one place - you invoke it instead of re-describing it.
Now break it on purpose. Build a case that violates it and follow the consequences until something visibly fails. Where does the failure first show up — and would you have noticed it if you had not been looking?
Hint: The dangerous rules are the ones whose violation still produces an answer. If yours fails loudly, try to find one that fails quietly.
Answer:
Nothing is saved. You retype the instructions, and small wording changes make the output drift week to week.
Section
Your garbage-in defense
Concept
This is the heart of what you asked for. Before we build anything, we fill this in. Six questions:
| Field | The question it answers |
|---|---|
| Goal | In one sentence, what should this tool do for me? |
| Inputs | What do I give it each time, and where does that come from? |
| Outputs | What exactly comes back, and in what format? |
| Constraints | What must it always do, and never do? |
| Success test | How will I know it worked - a clear pass vs. fail? |
| Reuse | Once, or repeatedly? (This picks Artifact vs. Skill vs. API.) |
Intuition
The worksheet isn't paperwork - it's a readiness test. Every blank you can't answer is a place the AI would have to guess, and a guess is exactly where garbage comes from.
So an unanswerable field isn't a failure - it's the worksheet doing its job, telling you the idea needs one more thought before it's worth building.
Counterexample
Discussion prompt
The worksheet isn't paperwork - it's a readiness test. Every blank you can't answer is a place the AI would have to guess, and a guess is exactly where garbage comes from.
That is stated as though it always holds. Do one of two things: produce a case where it fails, or say precisely what rules such a case out. "It just does" is not on the menu.
Hint: Hunt at the extremes first — zero, one, negative, empty, equal. If every extreme survives, the reason they survive is the proof.
Anomaly
Predict first
A student writes this, and it looks reasonable:
Skip the worksheet and ask straight away: "Build me a tool for my notes."
It is wrong. Say what breaks — and say it before you turn the page.
Correct: The AI fills the blanks with its own guesses - generic summary, wrong shape, made-up owners.
Fill the six fields first, then ask.
Why: The AI fills the blanks with its own guesses - generic summary, wrong shape, made-up owners.
Trap
Skip the worksheet and ask straight away: "Build me a tool for my notes."
No Goal, no Inputs, no Output format given
Why: The AI fills the blanks with its own guesses - generic summary, wrong shape, made-up owners.
You tweak the prompt over and over chasing the output you pictured
Why: Garbage in, garbage out - you never told it what 'good' looks like.
Fill the six fields first, then ask.
Goal, Inputs, Outputs, Constraints, Success test, Reuse all answered
Why: The spec is on paper, so the request is precise and there's nothing left to guess.
First build is close; you adjust once, not ten times
Why: Good inputs in -> usable tool out. The map did the hard part.
Two truths and a lie
Sort into buckets
Some of these hold up and some are the exact mistakes this lesson is built to prevent. Sort them.
Ranking
Put in order
Put the moves of Fill the worksheet, live: our meeting notes into the order they have to happen.
Why: These are the moves of the worked example in the order it makes them, and each one is set up by the one before it. One sentence, one job. If the Goal needs an 'and', it might be two tools.
Worked example
Watch all six fields get answered for the running idea. Notice how the tool almost designs itself once they're filled.
Goal: "Turn my raw meeting notes into a clean list of action items."
Why: One sentence, one job. If the Goal needs an 'and', it might be two tools.
Inputs: the pasted notes I type or dictate during the meeting.
Why: I named exactly what goes in and where it comes from - no ambiguity.
Outputs: a numbered list - each row = task, owner, due date.
Why: I specified the shape. 'A summary' would have been too vague to check.
Constraints: always include an owner; never invent a date that wasn't said.
Why: Always/never rules are how I stop the AI from filling gaps with fiction.
Success test: every action in my notes appears once, with an owner; nothing invented.
Why: A clear pass/fail I can eyeball in ten seconds.
Reuse: repeatedly, most weeks, in the Claude app -> a Skill.
Why: The Reuse answer picks the tool for me - exactly as the decision branch promised.
Reverse engineer
Discussion prompt
Work backwards. The example finished here:
Reuse: repeatedly, most weeks, in the Claude app -> a Skill.
What was it asked to do, and what must it have been given? Reconstruct the problem from its answer.
Hint: Every quantity in the result had to enter somewhere. Account for each one.
Answer:
Watch all six fields get answered for the running idea. Notice how the tool almost designs itself once they're filled.
Concept
Your only homework: jot down 2-3 things you wish you had a tool for - messy, half-formed is fine. We'll run one through this worksheet together, live, next time.
Bring whatever you get. A half-filled worksheet is still a huge head start - the gaps are what we'll work on together.
Ranking
Put in order
These are the steps of Fuzzy idea -> working tool, every time, scrambled. Put them back in order before the next slide shows you.
Why: This is the order the recipe itself gives. Recalling the sequence without the slide in front of you is the difference between recognising the method and being able to run it — most of what goes wrong in practice is a step done out of turn.
Pattern
This is the spine of the whole track. Any idea you ever have goes through the same four moves:
Map, check, pick, build. That order is your guard against garbage in, garbage out.
Edge cases
Discussion prompt
Fuzzy idea -> working tool, every time works on the cases you have just seen. Push it to the edge: what is the most degenerate input it still handles — empty, zero, one item, everything equal — and what is the first case where it stops being true? Name the case, not just "it breaks".
Hint: Try the smallest legal input, then the largest, then the one where two things collide. Methods are specified at their edges; the middle takes care of itself.
Answer:
This is the spine of the whole track. Any idea you ever have goes through the same four moves:
Elimination
Eliminate the wrong options
You want Claude to draft a reply every time a new message lands in your support inbox - automatically, with no one opening a chat. Which approach fits?
3 of these 4 are wrong. Strike them one at a time, and say what rules each one out before you strike the next. The survivor is the answer.
Survives elimination: A
Why: The job must run on its own, triggered by other software (the inbox), with no person in the chat. That 'runs automatically / inside another app' case is exactly what the API is for - which is why the track saves it for Session 4 and only if a project needs it.
Check
Use the two questions - how often, and where it runs. Decide before you tap.
Check your understanding
You want Claude to draft a reply every time a new message lands in your support inbox - automatically, with no one opening a chat. Which approach fits?
Answer: A
Why: The job must run on its own, triggered by other software (the inbox), with no person in the chat. That 'runs automatically / inside another app' case is exactly what the API is for - which is why the track saves it for Session 4 and only if a project needs it.
Elimination
Eliminate the wrong options
Someone says: "Make me a tool that organizes my emails." Using the mapping worksheet, what's the FIRST problem?
3 of these 4 are wrong. Strike them one at a time, and say what rules each one out before you strike the next. The survivor is the answer.
Survives elimination: A
Why: 'Organizes my emails' doesn't say what good looks like - sort how? into what? what's the output shape? With no clear Goal (and no Output or Success test), the AI must guess, and guesses are where garbage out comes from. Tighten the Goal first.
Check
Run it against the worksheet before you answer.
Check your understanding
Someone says: "Make me a tool that organizes my emails." Using the mapping worksheet, what's the FIRST problem?
Answer: A
Why: 'Organizes my emails' doesn't say what good looks like - sort how? into what? what's the output shape? With no clear Goal (and no Output or Success test), the AI must guess, and guesses are where garbage out comes from. Tighten the Goal first.
Connect it up
Draw it
One page, no notation unless you need it: draw how these connect — Know What's Possible · When to Use What · The Mapping Worksheet. Put an arrow wherever one of them is what makes another possible, and label the arrow with why.
Recap
| The move | The shortcut |
|---|---|
| Decide the tool | How often? + Where does it run? |
| Make an idea buildable | Answer all six worksheet fields |
| Test readiness | Any blank field = a gap to close first |
| Build well | Ask precisely, then judge by the Success test |
Next session: you bring 2-3 fuzzy ideas; we map one together and start turning it into your first real tool.
Want this taught 1-on-1? Alexander tutors AI Foundations — $55/session, free consultation.