A gentle on-ramp to AI

AI:
Start Here

Scott Hanselman
VP, Member of Technical Staff
Microsoft / GitHub
hanselman.com
Portland skyline with Mt. Hood
Scroll
01 — A Level Set

I'm an AI vegan.

I don't think AI belongs everywhere. I do not generate images or video.

But when it helps…

it can give one person the power of a team.
02 — What This Is

Not hype. A gentle on-ramp.

Not This
  • Hype
  • "AI will replace everyone"
  • Deep math
This Is
  • A gentle on-ramp
  • Real talk about what AI actually does
  • And how you can use it
03 — The Simplest Explanation

AI is

the most powerful
autocomplete ever built.
04 — Demo

Next-Word Prediction

The quick brown fox jumped over the lazy dog
AI predicts the next word.

Not because it knows the answer… because it has seen billions of sentences like this.

05 — What AI Actually Does

AI predicts patterns.

AI Predicts Patterns

It processes massive amounts of text and learns statistical relationships between words, concepts, and ideas.

It Doesn't
  • ❌ Know facts
  • ❌ Understand truth
  • ❌ Check reality

It predicts what sounds right.

06 — The Mental Model

How It Works

1 Prompt
→
2 Model
→
3 Prediction
→
4 Response

Pattern prediction at massive scale.

07 — A Very Portland Analogy

Remember the Portlandia sketch.

"Is this chicken local?"
Before eating it… they go visit the farm and learn the chicken's whole life story.
That's the energy we need for AI.
Is this chicken local? - Portlandia reference
08 — The Portland Way to Use AI

When AI gives an answer, ask the Portland question.

Where did this come from?
  • What patterns produced this answer? What model is this?
  • Who owns it, who trained it, what was it trained on?
  • Is it correct? Is it biased? Did I bias it? Did I ground it in facts?
09 — Because AI Can Be Wrong

AI sometimes predicts things that sound correct…
but aren't.

Confidence ≠ Accuracy
10 — Treat AI Like This

Think of AI as:

A very fast intern.
Helpful
Creative
Productive

But you still need to check the work.

11 — AI as Leverage

AI can act like:

🔍
A researcher
✍️
A copywriter
💻
A junior developer
🎨
A designer
💡
A brainstorming partner
📊
A data analyst

One person can now do much more.

12 — Chat vs Agents

Two Modes of AI

Basic AI

You ask → AI answers
One question, one response.
Like texting a really smart friend.
vs

Agentic AI

AI can take multiple steps.
It plans, executes, and adjusts.
Like delegating to a capable teammate.
13 — Why Founders Should Care

Starting a company used to require:

$$$
Capital
10+
Employees
Years
Time
AI changes the ratio. Agents work for you.
14 — The Shift Happening Now

AI should not replace builders.

It should multiply them.
15 — Why This Matters

Every generation gets a tool that expands opportunity.

Every generation panics.

Printing Press
Spreadsheet
Internet
Smartphone
AI

Now: AI

16 — The Opportunity

The people who win with AI
won't be the ones who trained the models.

They'll be the ones who learn
how to use them creatively.
17 — AI: Start Here

Your Next Four Steps

1

Ask questions

Don't just accept the first answer. Dig in.

2

Break problems into steps

AI works better with small, clear tasks.

3

Verify outputs

Check the work. Always.

4

Experiment

Try things. Break things. Learn.

You don't need to be an AI engineer. You do need to be curious.

Be curious. Be skeptical. Be creative.

And remember the Portland question:

Where did this answer come from?
Is this chicken local?
Scott Hanselman
hanselman.com · Scott Hanselman · Microsoft / GitHub