A primer · Alan Seideman

Working with AI

What these tools actually are, why they slow down, and how to get noticeably more out of them.

Applies to Copilot and Claude alike. The tools differ; the way they work does not.

TURN 1 TURN 16
One column per turn. The bright bar is what you just typed; the stack above it is everything re-read to answer it.
Opening

The words we'll use

None of this is complicated. The vocabulary is what trips people up.

  1. LLMThe engine under Claude and Copilot.
  2. PromptEverything you send, not just what you typed.
  3. TokenThe unit of measure. About four characters.
  4. ChatOne conversation, re-read in full each turn.
  5. ModelWhich version you pick. Bigger costs more.
  6. Context fileYour decisions, saved for the next chat.
  7. ProjectRelated chats, plus background they share.
  8. .md fileMarkdown — plain text with formatting.
  9. ConnectorA direct line into a system you use.
  10. MCPThe standard plug. Think USB.
What it is

What is an LLM?

The simplest way to picture it: your phone's autocomplete, scaled up enormously. It doesn't look answers up — it predicts them.

ON YOUR PHONE Running late? See you at the IT SUGGESTS THE NEXT WORD office meeting game It learned which words tend to follow which. SCALED UP, THAT'S AN LLM Not a stored copy of everything it has read. The patterns in how writing fits together. So it predicts what fits — it never looks anything up.

It keeps patterns, not pages — so it always produces something, even when it's wrong.

What it is

What is a prompt?

It's everything you send to the LLM — not just the line you typed.

WHAT ACTUALLY GETS SENT — EVERY TIME Project instructions and background Files and images you attached Everything already said in this chat The line you just typed YOU WRITE only this part

You write one line. All of it is the prompt.

How a chat works

Every message re-reads the whole conversation

There is no memory between messages. It starts from the beginning, every time.

OLDEST NEWEST TURN 1 1 message read TURN 5 9 messages read TURN 10 19 messages read
How a chat works

What is a token?

It's the unit of measure — everything read and written is counted in tokens.

ONE TOKEN ≈ FOUR CHARACTERS Pro posal doc re view five tokens, three words WHAT YOUR LIMIT COUNTS Tokens in everything it reads Tokens out everything it writes

Both are counted. A long chat spends most of its budget re-reading.

How a chat works

A file you attach is read again every turn

OLDEST NEWEST SPREADSHEET TURN 1 read once SPREADSHEET TURN 2 read again SPREADSHEET TURN 3 and again

You attach it once. You pay for it every turn — and a pasted screenshot counts too.

Pro tip

Keep each chat to one subject

Everything in the chat gets re-read — including the parts that have nothing to do with your question.

ONE CHAT, FOUR JOBS CLIENT A CLIENT B CLIENT C EXPENSES “Which client did you mean?” ONE CHAT, ONE JOB Straight to the answer. Shorter, cheaper, and nothing to confuse it.
Models

Bigger models think harder and cost more

Same kind of engine, different sizes. The cost is per token — in and out.

LIGHT MEDIUM HEAVY Claude Haiku 4.5 $1 in · $5 out quick rewrites, short emails Claude Sonnet 5 $3 in · $15 out the sensible default Claude Opus 5 $5 in · $25 out slow, costly, thinks hardest

Prices are per million tokens — so the model you pick moves the bill five-fold.

Pro tip

Match the model to the job

When you're doing thisReach forBecause
Reasoning across several documentsOpus 5Real judgment, and the answer has to hold up
Everyday drafting and questionsSonnet 5What most work actually needs
Quick rewrites and short emailsHaiku 4.5Quality isn't the constraint here
Follow-ups, once the thinking is doneDrop a levelDon't pay twice for the same reasoning

Being economical on the one task where quality shows is a false saving.

Pro tip

Keep the thinking, drop the bulk

CHAT ONE long and expensive summarise CONTEXT FILE decisions, mappings, what it got wrong attach CHAT TWO short, fast, already informed

Nothing new — it's the role and project reference docs we've always written, so the next person reads the memo instead of the whole history.

Pro tip

How to get a summary: just ask

KITCHEN REMODEL · COMPARING THE THREE CONTRACTOR BIDS “Summarise what we worked out here so I can reuse it — decisions, numbers, and anything you got wrong.” Here's a summary of what we decided: kitchen-decisions.md Markdown · 3 KB Save it. Attach it to the next chat — or drop it into the project. Reply to Claude…
Keeping what you learn

A project shares one file with every chat inside it

Claude PROJECTS Kitchen Remodel Q3 Board Deck Home Insurance RECENTS Contractor bids Tile + backsplash Permit questions Kitchen Remodel One place for everything about the remodel. PROJECT KNOWLEDGE kitchen-decisions.md Markdown · 3 KB · from your last chat Written once. Read by every chat below, automatically. CHATS IN THIS PROJECT Comparing the three contractor bids Backsplash tile pricing for 32 square feet What permits does the county need?

Every new chat starts already knowing the background. You stop re-uploading.

Keeping what you learn

Markdown is just a few symbols

On the left, what you type. On the right, what those symbols turn into.

What you type
# Kitchen Remodel ## Decisions - Shaker doors, white - Quartz counters **Budget: $32,000**
What it becomes

Kitchen Remodel

Decisions

  • Shaker doors, white
  • Quartz counters

Budget: $32,000

Plain text that still knows what's a heading and what's a list — nothing to learn beyond a handful of marks.

Keeping what you learn

Save working notes as markdown

Exactly the same notes, saved four ways. The bar is the file size — what the tool has to read through before it reaches your words.

notes.md 6 KB notes.txt 6 KB notes.docx 42 KB notes.pdf 138 KB
Putting it together

What this looks like in practice

Claude PROJECTS Kitchen Remodel Q3 Board Deck Home Insurance RECENTS Cabinet options Tile + backsplash Permit questions Contractor bids Kitchen Remodel Everything about the remodel lives in here. PROJECT KNOWLEDGE budget.xlsx decisions.md measurements.md quotes.md Not just notes — a spreadsheet works too. Every chat reads all four. CHATS IN THIS PROJECT Shaker vs slab doors — cost and upkeep Backsplash tile pricing for 32 square feet What permits does the county need? Comparing the three contractor bids

Now swap kitchen remodel for a client. Same structure, exactly.

Next

Three things to try this week

CHAT ask .md
01

Summarise a chat into a file

Ask for the decisions, the numbers, and what it got wrong. Save the .md.

PROJECT
02

Group related chats into a project

Gather everything about one piece of work in a single place.

PROJECT SHARED FILE
03

Add context every chat shares

Drop that file — or a spreadsheet — in. New chats start already knowing it.

Going further

Advanced

Where this is heading — connecting the tools to your systems, how Copilot and Claude actually differ, and working straight off your own files.

Advanced

Connectors — one plug into everything

A direct line from the tool into a system you already use. MCP is the standard plug shape that makes it possible — think USB.

Microsoft 365 SharePoint Outlook Notion Gmail Your files Claude reads from the source
Advanced

Copilot is the machine. Claude is the brain.

Copilot wraps a model — the brain — and connects it to your Microsoft 365 through a layer Microsoft calls Work IQ.

COPILOT MODEL SLOT Claude …or GPT, or another model WORK IQ Microsoft's layer into your email, files, meetings and chats Word Excel Outlook Teams SharePoint

Microsoft now runs Anthropic's models inside parts of Copilot — so the brain in the slot may be Claude either way.

Advanced

Claude chat vs Cowork

CLAUDE CHAT IN THE CLOUD CHATS FILES Chats and files live in Claude's cloud. COWORK IN THE CLOUD CHATS FILES + ON YOUR COMPUTER Files & folders on your machine — a second layer Cowork reaches.

Chat works over what's in the cloud. Cowork also reaches the files and folders on your machine.