Hey, Ivy here.

Custom GPTs get messy when you try to make them do everything.

The better move is narrower.

Give the GPT one role.

Give it one job.

Give it a format.

Then give it examples or reference files that show what good looks like.

Do not start with:

be my all-purpose assistant.

Start with something specific, like:

turn these call notes into a warm follow-up email with next steps and open questions.

That is easier to test.

Easier to improve.

And easier to trust.

The clearest GPT wins.

Not the biggest one.

Tomorrow, I will show you how to talk to your PDFs instead of hunting through them.

Hey, Ivy here.

ChatGPT memory is useful when it saves real preferences.

It gets weird when stale context hangs around too long.

The fix is simple.

Treat memory like settings, not like a diary.

Save durable things, like how you like answers structured, recurring context, or format rules you use all the time.

Be careful with temporary projects, old drafts, and preferences that changed.

Then ask:

What do you currently remember about how I like to work?

Show me a short list so I can keep, remove, or correct anything.

That one check keeps memory helpful instead of strange.

Tomorrow, I will show you how to make custom GPTs useful without the fluff.

Hey, Ivy here.

If your spreadsheet is slowing you down, use AI for the annoying first pass.

Here are five prompts for spreadsheets that are actually useful.

Ask it to explain a spreadsheet formula in plain English.

Ask it to write a spreadsheet formula for the exact result you want.

Ask it to spot weird patterns or outliers in the spreadsheet.

Ask it to clean inconsistent spreadsheet data.

And ask it to turn the spreadsheet into a short summary you can send to someone else.

That is the useful framing.

Use AI like a spreadsheet assistant, not the spreadsheet owner.

Keep your judgment.

Let the model handle the friction.

Tomorrow, I will show you how to use ChatGPT memory without letting it get weird.

Hey, Ivy here.

Search smarter by splitting the job in two.

Use AI search when you need a quick summary, a comparison, or the shape of a new topic.

Use Google when you need the original source, the freshest update, or exact details.

Best workflow:

ask for the summary,

open the cited links,

then verify what matters.

The smooth answer is not the proof.

Use AI search to get oriented.

Use Google to inspect the evidence.

Tomorrow, I will show you five spreadsheet prompts that turn AI into a useful first-pass spreadsheet assistant.

Hey, Ivy here.

If an AI answer feels vague, the prompt may be the problem.

Try this simple structure:

Role. Task. Format.

Role tells the model who to be.

Task tells it what to do.

Format tells it how to hand the answer back.

For example:

You're a skeptical editor.

Review this draft for clarity.

Give me five bullet points and rewrite the worst paragraph.

That is usually enough to make the result sharper.

Don't make the prompt longer.

Make the request easier to follow.

Hey, Ivy here.

If you keep re-explaining the same project every time you open Claude, Projects are finally useful for one simple reason.

They remove setup.

Projects let you keep the chats, instructions, and reference material for one recurring workflow in one place.

That makes them a good fit for things like proposals, content series, launch plans, and weekly reports.

Start simple.

Give the Project one standing instruction:

When I paste new material, look for gaps, contradictions, and missing decisions first.

That is enough to make the next chat better.

Projects are not magic.

They just stop useful context from resetting every time.

Hey, Ivy here.

Before you send a draft, try one fast AI check.

Paste your email, post, or update into AI and ask:

What is unclear, easy to misread, or missing context here?

That is a better prompt than just asking it to make the draft sound better.

You want it to read like someone seeing the message for the first time.

Have it flag vague lines, hidden assumptions, and the question the reader would still have after reading it.

Then fix those spots and keep your own voice.

Clarity is usually a better upgrade than more polish.

Hey, Ivy here.

If your inbox feels like a wall of random priorities, here’s a better way to start.

Paste your recent emails into AI and ask it to sort them into three buckets: urgent replies, important but not urgent, and FYI only.

Then ask for the top five actions you should take next, plus anything time-sensitive or waiting on your reply.

That won’t magically do your inbox for you.

But it will help you see what matters before the morning disappears into reactive reading.

If your inbox is overflowing, don’t start by reading more.

Start by summarizing smarter.

Don't ask for one draft. Ask for five — bold, conservative, weird, plain, contrarian. Then pick. Selection beats generation.

Ask AI: what's missing from this? It surfaces the question you forgot to ask and the angle you forgot to consider.