Hello, object store: it's just a bucket

An object store is a flat map from a key string to an immutable blob of bytes that you PUT once and GET back forever — and 'forever' and 'infinitely' are promises the machinery underneath has to earn.

Scene 01

Hello, object store: it's just a bucket

Watch
Try it
Predict
Capture

What to watch for

A client stores some bytes under a name, then asks for that name back later. Watch the box on the right fill up, then hand the same bytes back. That box is the whole product — for now it looks like magic.

A client on the left sends two requests into one box: PUT(key, bytes) writes some bytes under a name, GET(key) reads them back.
The box is a bucket — a flat list of (key → bytes) tiles, never a folder tree.
Peeking inside cracks it into two faint planes that the rest of the curriculum will fill in.

Implementation

Highlighted lines are the ones running in the diagram right now.

Client.put

write some bytes under a key into a bucket

1# store bytes under a name; no folder is created
2s3.put_object(bucket, key, bytes)
3# key is one flat string: "logs/2026/app.log" != a path

Client.get

read exactly those bytes back by key

1# returns the exact bytes last stored under key
2bytes = s3.get_object(bucket, key)

Sources

Where this sits in Build an S3-style distributed object store

Scene 01 of 12, in the The stakes act — What an object store promises — and the durability problem.. Bucket, object, key, PUT, GET — the flat key→bytes model, before we earn the word "forever."

Up next. If 'forever' is a promise and not a fact, the first thing it has to survive is the hardware underneath — and that hardware is millions of disks that are failing right now.

All 12 scenes in Build an S3-style distributed object store · Every curriculum

Built with Arqly

Every scene in Build an S3-style distributed object store builds on the one before it.All 12 Build an S3-style distributed object store scenes