At the club today

Hi, Red here. In Coding First Steps, we looked at how screens and code fit together. This time, a question for AIKO got me wondering about something else.

“AIKO, when’s our next game meetup?” We decided the time ourselves. Would AI know it? If our club notice is not on the public internet, AI needs a way to look it up.

So Nea, Kirari, and I decided to find out how an AI wiki that looks up club information works. We’ll use a “club game archive” containing game guides and meeting notices. The archive and schedules are fictional examples.

Meet the idea in a comic

A wiki collects and connects information

Here, a wiki means a place where we organize articles by topic and link related articles together. It does not have to be open to everyone like Wikipedia. A wiki can be for club members only.

An AI wiki connects this material to AI: it finds passages related to a question and lets AI refer to them when answering. Putting a file in an archive does not automatically make AI know everything in it. The material must be ready to search, and the reader must have permission to access it.

Look at the notice instead of relying on memory

Kirari

To know our schedule, it needs to read our notice first!

The material AI learned from earlier may not include a schedule we just agreed on. Instead, we can first find relevant information, then give it to AI to help write an answer. This is called retrieval-augmented generation, or RAG. For now, think of it as “looking up information and using it to answer.”

1. Ask a question: “When’s our next game meetup?”
2. Find a source: look up the “Game Meetup Notice.”
3. Write an answer: use the notice’s “Friday, 4 p.m.”
4. Check the original: open the cited source and check its date and conditions.

Four new words: start with what they do

These are not necessarily four steps that run once in this exact order. Ingestion may include splitting text, creating embeddings, and building indexes. A consolidated article may be ingested again. Today, just remember that the four roles are different.

Managing information and finding it

As the archive grows, we can separate two jobs: collecting material and managing its review status, and finding relevant material for a question. Both help new information reach an answer.

“Searchable” and “public on the internet” are different. An article restricted to club members must keep that restriction when it is searched.

Watch the short · Korean audio

Three small exercises with Red

No software to install. Choose an answer or write one short line in a note.

1. 1. Pick the source-based answer

   The notice says “Friday, 4 p.m.” Which answer uses it? ① “We usually meet on Saturdays, so probably Saturday.” ② “According to the notice, Friday at 4 p.m.”

   Show the answer and why

   ②. It uses the time in the original notice. Also check the notice’s date and any later changes.

2. 2. Choose a relevant article

   For “question → find relevant material → answer using it,” which article should we check first: “Game Meetup Notice” or “Game Wallpaper Collection”?

   Show the answer and why

   Start with “Game Meetup Notice.” A wallpaper article may mention games, but it is unlikely to tell us the meeting time.

3. 3. Check who may read it

   The archive contains a club notice and private contact details. Should every member see both?

   Show the answer and why

   No. Check access for each document. Being in the wiki does not make a document readable by everyone.

Before we close the archive

Today we explored answering with sources and the four supporting roles. Next, we’ll find out why searching for “games to play together” can return an article about two-player co-op games.