Claude Code Basics — Clearing Context (lesson deck)

Claude Code Basics · the habit that keeps you fast

Clear your context. Often.

Every message stays in the conversation. The longer it runs, the more the model re-reads — so a giant chat quietly works against you.

🐌

Slower & pricier — it re-reads the whole history every single turn.

🧶

It loses the thread — old, stale details bleed into new answers and it contradicts itself.

🎯

It drifts off-task — five topics ago is still "in the room," muddying focus.

Rule of thumb: context is RAM, not storage — keep the working session short.

The right way · don't just wipe it

Checkpoint, then clear

1

Save the durable stuff first

Ask the AI to write decisions, specs & next steps into .md files + memory. Files are your long-term memory.

2

Clear the context

Run /clear to start a clean session. Nothing's lost — it's on disk.

3

Reopen pointed at the files

"Read those notes & the memory, then tell me where we are." Short context, full knowledge.

The mistake is clearing without saving. Write to disk first — then wipe freely.

Steal this · the checkpoint prompt

One prompt before you /clear

Before I clear this conversation, act as a project archivist. Review everything we decided and built, and persist it so nothing is lost when context is wiped: 1 Save each durable decision / spec / reusable asset into a clearly-named .md file (distill conclusions, don't dump the chat). 2 Update my memory/notes with one-line pointers to those files. 3 List any OPEN threads + next steps as a short TODO. 4 Tell me what's already safe elsewhere (committed code, docs). Then give me a one-paragraph "where we are" I can paste into a fresh session.

Treat context like RAM, not storage.

Write decisions to disk · keep the chat short · clear often.