Women in Product Nordics / April 2026

Claude Code for the Curious

What happens when you tell your computer what to build and it actually does it

Olga Safonova

Matrikel1, Copenhagen 15 April 2026

The Basics

What is Claude Code?

You describe the outcome. It figures out the steps.

2

Live Demo

Let's explore
something together

Pick one. I'll type it in. We'll see what happens.

Explain a repo. Point it at a public project, get the full picture
Analyze our attendees. Turn tonight's registration CSV into a visualization
Edit this deck. Add a slide to this presentation, live

3

Beyond Code

Things you can ask it

"What does this project do? Give me a summary"
"Compare these two competitor landing pages"
"Find every place we mention pricing in our docs"

"Write user stories from this requirements doc"
"Analyze this spreadsheet and flag outliers"
"Draft release notes from this list of changes"

Be specific about what you want. Vague asks get vague results.

4

Demo

What I Actually Use It For

Morning standup prep

One command. I type /today and it builds my daily ToDo from DevOps, GitHub, Granola meeting notes, and session history.

Demo shown live at the event

The recording is not published. It was captured on a work machine and shows internal repositories and colleagues by name.

Runs on my work network, so this one was pre-recorded for the room.

5

Demo

What I Actually Use It For

Colleague asks a question

I type /lmwtfy and it researches the answer across source code, wiki, Help articles, and DevOps.

Demo shown live at the event

The recording is not published. It was captured on a work machine.

Searches source code first, then the customer-facing help center, then wiki and DevOps. Saves the answer and proposes wiki updates for knowledge gaps.

6

Live Demo

What I Actually Use It For

How I actually work with AI

I'll type /insights right now. It reads 30 days of my session history and generates:

Which projects I worked on and how much time each got
My interaction style: how I prompt, correct, and iterate
Auto-generated rules from patterns in my behavior
Recommendations based on my actual usage

Runs locally. No data leaves my machine.

7

What I've Learned the Hard Way

It invents facts with total confidence

Fabricated numbers in my slides. Faked article summaries. Called real tools fictional.

You are the quality filter

It won't tell you when it's wrong. If you don't check, you ship confident garbage.

It gets stuck in loops

Same failing approach, five times in a row. You have to notice and interrupt.

You own what it builds

Writes a feature in two minutes. When it breaks at midnight? That's your problem.

It's incredible anyway

I've shipped tools used daily by my team, automated hours of weekly reporting, and built integrations that would have taken months. The tradeoffs are real, but I wouldn't go back.

8

Your Turn