have you ever written a program?

1. yes, quite a lot
2. a little, at school or for fun
3. never

how often do you use AI chats like ChatGPT?

1. every day
2. every week
3. now and then
4. never

have you ever worked with an AI agent, like Claude Code, Codex or OpenCode?

1. yes
2. I have heard of them
3. no, never heard of them

how well do you know what happens inside a computer?

1. very well
2. roughly
3. hardly
4. not at all

how many different values can one byte hold?

1. 8
2. 16
3. 256
4. no idea

a photo on your phone: roughly how many numbers are stored for it?

1. a few hundred
2. a few thousand
3. tens of millions
4. no idea

at the lowest level, what does a computer work with?

1. words
2. pictures and sounds
3. numbers: zeros and ones
4. no idea

you switch off your computer. what happens to what is in its RAM?

1. it stays there
2. it is lost
3. no idea

how important will it be for your career to understand AI and to work well with it?

1. essential
2. very important
3. somewhat important
4. not important

why this module

AI answers any question in seconds.

it writes, summarises, translates and explains. it seems to know everything.

or does it?

so why learn anything at all?

the same question, fifty years ago:

everyone has
a calculator.
why still
learn maths?

image: AI-generated (gpt-image-2.5-flare)

why maths, if you have a calculator?

to write the problem down so that the calculator can solve it.

to notice when 4,000 comes out and 40 would be right.

to know what it cannot calculate at all.

the same holds for computers, and for AI.

a computer is just a bigger calculator.

it turns numbers into numbers, billions of times a second. to judge what comes out, you have to understand what it does.

image: AI-generated (gpt-image-2.5-flare)

pilots still learn how a plane flies.

the better the autopilot, the rarer the moment a human takes over, and the more that moment depends on understanding.

image: AI-generated (gpt-image-2.5-flare)

the most powerful calculator humans built so far

it answers anything in seconds.

it sounds right, even when it is wrong.

it cannot check its answer against the real world.

you learn to work with it: cut, check, measure.

image: AI-generated (gpt-image-2.5-flare, edited with gpt-image-2.5-sunburst)

and to understand AI

what comes next

this module is the foundation. the next one builds on it.

would you like to have a better understanding of computers?

1. yes, definitely
2. a little
3. not really

what you will learn

the map lives on the website

why a project,
and not just
lectures?

image: AI-generated from photos of the real device (gpt-image-2.5-sunburst)

why a project?

learn by doing

concepts stick when you have built them, broken them and fixed them yourself.

problems first

every problem meets you on your own device before it gets a name in class.

reality answers

a device does not care how good an answer sounds. it works, or it does not.

the goal of the lifi project

there is a photo
on this laptop.
how does it get
to the other one?

image: AI-generated (gpt-image-2.5-flare)

the device

a lamp and an eye on the front of each device. two devices face each other, so light travels both ways.

why a lamp

everything interesting in a computer is invisible.

this project makes it visible: a few times a second, on your desk, where you can explore it with your own hands.

image: AI-generated (gpt-image-2.5-flare)

the journey of one photo

every stop on the way is a concept of this module.

the rules

take a pen and a piece of paper.

the screen is the lamp. you are the sensor.

write down every symbol you see, using the alphabet at the bottom.

round 1: four colours, slowly

start (s)

sent

round 2: eight blues

start (s)

sent

round 3: faster

start (s)

sent

round 4: where is the message?

start (s)

sent

round 5: did everything arrive?

start (s)

sent

what went wrong?

→ challenge 1: the alphabet

→ challenge 2: the word

→ challenge 3: the listener

→ challenge 4: the packet

one new difficulty at a time

each challenge adds exactly one new difficulty. everything else stays as you know it.

what every challenge produces

measurement logs

what did you change, what did you keep the same, and what came out?

a description of your solution

enough for another team to rebuild it. from challenge 3 on, a full protocol specification.

a log of mistakes

where the assistant suggested something, and a measurement proved it wrong.

nothing of this is handed in or graded. it is for your team.

how we work

two students, two devices

one device each, so you own both ends of the link. split the work, and swap now and then.

the lamp has the last word

your assistant knows nothing about your device. measure before, measure after, and note where it was wrong.

try the demonstrators

a small toy in the browser for almost every concept: switch a byte, break a file, add noise.

fourteen mondays

1

kick-off: overview & setup

devices, setup, challenge 0 starts

2

cutting problems, algorithms and programs

challenge 0: your first program

3

measuring and experimenting

challenge 0 check

4

analog and digital, symbols and information

challenge 1 starts

5

signal and noise

competition: challenge 1

6

number systems and code systems

challenge 2 starts

7

sampling and synchronization

challenge 2: faster

8

throughput and limits

competition: challenge 2

9

protocols

challenge 3 starts

10

logic and arithmetic

interop test

11

memory and storage

competition: challenge 3

12

errors and redundancy

standardisation session, challenge 4 starts

13

compression

challenge 4: faster and shorter

14

abstraction and layers

final: challenge 4

a typical monday, 15:00 to 18:15

15:00	look back: where are you stuck?
15:30	input: one concept
16:00	experiment, all together
16:30	break
16:45	workshop
18:00	outlook and your task for the week

every monday, the longest block is yours: you build.

every monday

room HD0001

mondays, 15:00 to 18:15.

laptop and device

bring both, every time. charge the laptop before you come.

attendance list

sign it every week. it is for statistics, not for grades.

the exam

60 minutes

multiple choice on the concepts, everyone on their own. the project is where you meet them.

+5 % bonus

if all five of your challenges are accepted. all or nothing.

01234

practise

with the questions on every concept page. same format as the exam.

where to find what

docs.lifi-project.de

everything to read: concepts, challenges, setup

lms.hs-osnabrueck.de

announcements, changes, the schedule

your course folder

your assistant and your code

lifi.uber.space/cockpit

your link, live

demos.lifi-project.de

the demonstrators

how do you feel about this module?

1. looking forward to it
2. curious
3. a bit worried
4. worried

by the end of term,
a picture crosses that gap.

and along the way,
more than one light will go on.

image: AI-generated (gpt-image-2.5-flare)