When you can't see what you're too close to see
"What are my blind spots here? What am I not seeing that I should be?"
Fire it before a big decision or hitting publish - a "check my thinking, not just my work" one.
Content. Copy. Compelling Brand Strategy.
You're in! π
Everything's inside. Some tiny upgrades, two seconds each, mighty on what Claude hands back.
Prefer to read offline?
Download as PDFSteal one prompt, make one upgrade, do it today. That's all I want for you. {The yes-man one is a great place to start π}
hey, i'm carinda!
Your marketing + online business bestie - helping you get seen, loved, selling + scaling {even when you don't}.
I run a multi-seven-figure online business with no full-time team - with AI making it all lighter, smarter and faster, and my best human bits kept for where they matter most.
that's me x
Before we start
There's some really big stuff you want to be doing in Claude - but the small stuff should never be overlooked.
If you're using Claude, you might already know the most important parts are your skills, Projects, connectors, plugins, and knowing the difference between Claude, Cowork and Code. But there are also so many little things that uplevel the whole system - the ones that make it work with you, for you, as you, and be a better business collaborator altogether.
In this guide I'm sharing a handful of the ones I find the most powerful. Take what you need. β₯
One · Honest feedback
{'cause honestly, mine used to - until I did this too}
Okay so - you share an idea, and Claude says it's brilliant. Share kind of the opposite idea an hour later? Also brilliant. π« You ask "is this any good?" and it just gushes. And somewhere in your gut you're like, hmmm, I don't actually think you're being honest with me here.
AND IT'S NOT. Not all the time, anyway. Claude's one of the best models at giving real, constructive feedback - but it's not always automatic, and you can teach it to be so much better. Here's exactly what I taught mine {it lives in my brain and loads every time I open Claude}:
"Don't be a yes-man. Challenge my thinking, surface my blind spots, and tell me when I'm wrong and why. Give me a recommendation, not just options. If I push back and there's context to show I'm right, change your mind cleanly. I'd rather you be honest than agreeable."
And where you pop it so it actually sticks:
Teach it once, and honest is just how your Claude talks to you from here. β₯
Two · The way you ask
{this is the other half of getting the truth out of it}
Here's the thing most people miss: how you ask changes the answer you get. Claude wants to be helpful - so if you lean the question one way, it'll lean the answer right back at you.
Ask "is this any good?" and it'll hand you the good points. Ask it to "tear this apart" and it'll give you only the bad. Neither one is the honest picture - you've just told it which way to fall before it's even looked.
So don't hand your work over leaning one way. Ask for the whole truth instead:
"Can you assess this critically for me - the real pros, the real cons, and any gaps I might be missing?"
Now both sides are on the table. That's Claude being an actual advisor - not something that just tells you what it thinks you want to hear. β₯
Three · The one-liners
Sometimes one little line completely changes what Claude hands back. Tiny to type, massive difference. Swipe every one:
"What are my blind spots here? What am I not seeing that I should be?"
Fire it before a big decision or hitting publish - a "check my thinking, not just my work" one.
"I'm not technical, and I'm trying to understand [X] about AI/Claude. Explain it in plain English - like I'm smart, but brand new to this?"
Claude will go as jargon-free as you need - you just have to ask. No more nodding along to stuff you don't actually follow.
"Don't start yet - first, ask me any questions about gaps you can see, or things I might not have outlined for you."
Instead of guessing at half a brief, it interrogates the goal with you first. You'll be amazed what it flags.
"Do [X] for me - but don't just hand me the finished thing. As you go, explain HOW and WHY, and teach me the parts I could actually learn from - so I get the work done AND get smarter, instead of just handing my brain over."
It's so easy to let AI do everything and slowly go rusty. This flips it: you still get the work done, but come out sharper every time, not more dependent.
"Before we finish - look back over this whole conversation and pull out anything genuinely worth keeping: new rules I set, preferences I revealed, things you learned about how I work. Be selective. Then propose the exact updates I should make to my brain {or CLAUDE.md} for me to approve - don't change anything on your own."
Selective is the key word - you don't want it stuffing your brain with every random thing. You want it to propose, so YOU stay the one who decides what makes the cut. This updates your brain in real time - and remember, having a Claude brain is one of the most important things you can do. If you're in Claude AI, I recommend making this into a skill so it runs every single time.
Four · Talk, don't type
{this one completely changed my speed}
Okay, this is one of my favourites, and it's so simple. Most people type careful, tidied-up prompts at Claude - like they're writing something formal. It's slow, and it means Claude only ever sees the polished version of you.
So instead, I just talk to it. Not normal voice-to-text either {more on the tool below} - I give Claude a full stream of consciousness, messy and unpolished, exactly as it's coming to me.
Here's why that matters more than it sounds: when you talk to it like that, Claude gets to see the real you - your actual thoughts, before you've tidied them up. It's not only faster, but the real magic is that it gets to see how you actually think. I'll be talking through something, and halfway I'll go "actually - no, wait, it's this way instead." Because Claude saw that shift happen live, it genuinely understands my reasoning. It's not just doing what I told it. It's thinking WITH me.
Hand on heart, talking to Claude like this has 10x'd my speed in my business.
The tool I use for this is Wispr Flow. I use it across my whole business, not just with Claude - it cleans up my grammar as I go, saves everything to its own backend, and works on any device. Want to try it? Here's a free month β wisprflow.ai/r?CARINDA1
Five · Save your usage
Each time you send Claude a message, it looks back over the whole chat so it knows what you're talking about. The longer the conversation gets, the more there is to look through - and that's what eats through your usage faster. If a chat gets really long, Claude even shortens the earlier parts into a summary to keep up. So the fix is all about keeping your chats lean:
Onto something unrelated? Open a new chat - no history to re-read, so it's the leanest, cheapest work it'll do. One endless mega-chat is one of the biggest credit-drains going.
Don't want to lose a long chat? Get Claude to write a handoff summary, then paste it into a fresh chat:
"Write me a handoff summary I can paste into a new chat to pick up exactly where we left off - the goal, key decisions and why, where we're up to, and what's left."
Claude gone the wrong way? Don't type a long "no, not like that" - that just makes the chat longer AND muddier. Instead, hover your original message, edit it, and re-send. Claude picks up cleanly from there, like the wrong turn never happened - lighter on usage, and a better answer too.
{The real credit-saver - and what I teach.} Keep reopening the same long chat because it knows how to do something? Get Claude to turn it into a skill - just the instructions, none of the messy back-and-forth. Build once, and unlike a Project {locked in claude.ai}, a skill is portable across Claude and Claude Code.
The big models {Opus, Fable} are for proper thinking but use your usage faster. For quick stuff - reword this, tidy that - drop to a lighter one {Haiku or Sonnet} and save the big brains for when it counts.
Six · Connect it up
So many people are using Claude, but still doing half the work themselves. Copying and pasting things in. Finding the info manually first. Then carrying Claude's answer back to wherever it needs to go.
Here's the upgrade: a lot of the time, you don't even need to hand Claude the information. You can just tell it to go and get it. Once you connect Claude to the tools your business runs on, you can say things like:
And it justβ¦ goes and does it. No fetching, no carrying, no copy-paste. The ones you can connect in a couple of clicks {claude.ai β Settings β Connections}:
{Already got those connected? Gold star - the next page is where it gets fun.}
Seven · Custom connections
You'll find loads of connectors in Claude's own library - but your options don't end there. Some tools offer their own custom connectors. You won't find Meta in the library right now, for example, but google "Meta Claude MCP connection" {MCP being the bit that lets you plug Claude into the tool} and you'll often find you can connect it manually - just head to their help guide.
β οΈ Only connect the tools you're genuinely comfortable connecting, and know what each one can see and do. You're always in charge - Claude only ever touches what you let it.
One thing that trips people up: for these everyday "go grab this" jobs, a direct connection beats firing up Cowork every time. Cowork's brilliant for big, multi-step jobs Claude runs on its own - but it spins up a whole session and loads everything first, so it's slower and burns more credits. For a quick "read this email and tell me what they want," a connection just does it - faster, cheaper, done.
{And once you're deeper in - this is where my Claude Code world lives - you can connect it to basically anything: your ad account, your finances, even your whole website. But start with the ones above.}
Eight · Your dream team
Instead of one Claude doing everything, you build a team - a set of skills, each trained to do a single job brilliantly, in your voice. Each "hire" is built once, never forgets how you like things, and costs nothing to keep on. Here's my actual marketing crew - a bestie for basically every corner of it:
Ideas & Research
Brand Message
Social Media
Email & Lead Magnets
Copywriting
Other Content Marketing
Design
The clever bit? A skill can be handed from one person to another - so you can build your own crew from scratchβ¦ or just have mine, done for you, inside my club. And the biggest of them all - Maggie, inside the Lab - reads what's actually working on Instagram right now and writes ALL your content as you. π
Want the whole crew, done for you?
It's all in the clubNine · Feed it right
Here's a good one - once this clicks, everything you feed Claude gets better. Whenever you're uploading a document to Claude, give it as a Markdown file wherever you can.
"A Markdown what?" - don't worry, it's simpler than it sounds. A Markdown file {.md} is just a text file Claude really likes - clean and simple, and it still holds all your formatting. And you don't need to be techy: in Google Docs, just File β Download β Markdown (.md).
So anytime I'm handing Claude something - transcripts, ideal-client research, a big set of instructions - I give it as a proper .md file. Two reasons:
That one little swap - PDF to .md - quietly makes everything you hand Claude land better. β₯
Ten · Save the best for last
Here's one more that makes a real difference, and it's all about where things go. Claude pays the strongest attention to what's clearly structured and near the end of your message {that's Anthropic's own guidance}. So give it all your context first, then close with a quick recap of your non-negotiables:
"Quick recap before you start - you must ALWAYS: 1) [x], 2) [y], 3) [z]. And always write in my voice - re-read the voice section first."
Same info, but the bits that matter most are the last thing it reads - so they're the bits it actually holds onto. A tiny move, and it's the difference between Claude half-remembering your must-dos and nailing them every time. β₯
Eleven · Build it into a skill
When you find yourself doing the same kind of job again and again, that's a skill waiting to happen - and Claude has a built-in tool that builds it with you. It's called the skill creator.
Go to Customise β Skills and look for /skill-creator. Bring it into any chat, and when you want to turn what you've just done into a saved skill, simply ask:
It makes sure the work is captured in the very best way for skill creation - none of the messy back-and-forth, just the clean instructions. And the bonus: bring /skill-creator in any time you want to edit a skill too, and it'll supercharge the one you've already got. β₯
Twelve · Show, don't tell
{the technical name is few-shot prompting - but "show, don't tell" is all you need}
Here's one that changes everything, and it's so simple: when Claude isn't giving you what you want, stop describing what you want and show it instead.
Say it writes you a caption and it's just... not your tone. The instinct is to go "no, make it friendlier, that sounds stiff." But that's telling it - and it's left guessing what "friendlier" even means to you.
So show it. Rewrite one yourself, the way you'd actually say it, hand it back, and say:
"This isn't quite right - it doesn't sound like me. I've rewritten it below the way I'd actually say it. Study what's different between your version and mine, tell me what you notice, and use that from here."
Now it's not guessing. It's working from a real example of you. This is a proper technique - Anthropic's own top tip for better answers - and one good example beats a paragraph of instructions every time.
One to know: Claude picks this up within that chat. To make it stick for good, save what it worked out into your brain or memory {remember the wrap-up prompt back in point three π}. That's how a one-off becomes permanent.
before you go
Because it can be so much more than that. It can be a real force in your business - one that works for you, as you, and quietly takes the heaviness off your plate.
That's the part I care about most. Not the productivity or the output - the way it feels when everything doesn't land on you anymore. When you get to be the best, most human version of you inside your business, and actually love it again. More and more, not less and less. That's what all these tiny upgrades are really for. β₯
cheering you on x
When you're ready for more than two seconds...
Want to nail the big stuff?
Not certain you've fully nailed skills, connectors, plugins, Cowork and automations - or you haven't built your Claude brain yet? That's the exact thing I walk you through, step by step, in my Claude starter system.
Get the Claude Starter SystemWant Claude AND your marketing handled?
Want your Claude sorted AND your whole marketing handled by AI - the get-seen, get-loved, get-selling path, with the full dream team done for you? That's my club.
Come see the clubthis is co.
thisisco.com.au · @carinda.gosling.co · hello@thisisco.com.au
this is co.
Based in Australia, helping small biz owners right across the globe. Getting you seen, loved, selling + scaling.
Copyright © 2026 · Carinda Gosling | This is Co.