I thought I had Claude figured out: prompt templates saved in neat folders, a fat about-me doc I pasted into every single new chat. Then a post from an AI professional opened with one line that stung. 'You think you've mastered Claude,' and it turned out most of my tidy setup was extra effort producing worse answers, and I had no idea.
The creator lays out seven daily habits that hold you back, each one paired with a fix you can run today. A few are simple tool swaps: voice dictation at 150 words a minute, Connectors wired into Gmail, Slack and your meeting notes. The rest are mindset flips, like refusing to polish the perfect prompt and letting the model interview you until failure is off the table.
What struck me most was simple. None of these fixes need any technical skill, just small changes to how you talk to the tool. Here is my full breakdown of all seven, with the mechanism behind each one so you know why it works.
In partnership with
Discover laptops with built-in intelligence.
Shift your workforce into hyper-productivity mode with Dell Pro laptops powered by Intel® Core™ Ultra with Intel vPro®. Shop on Dell Premier.
*Ad
Your folders are poisoning the context
Most of us hoard reference docs, old drafts and past outputs, sure that more material means better answers. The author says delete them all. Folders leak stale work into the context, and the model starts copying yesterday's output instead of thinking fresh.
The fix is a setup with two parts, Skills and Projects. Nothing else. Skills hold reusable instructions; Projects hold the context for one specific piece of work.
I felt this one myself. Every time I dumped an old draft into a project, the next answer rhymed with the old one instead of improving on it. Fresh context is a feature, not a gap to fill.
Here's the test I now use: if a file isn't something the tool needs every single time, it becomes a Skill or it goes. Clean beats comprehensive.
Write the lazy prompt instead
This one surprised me most. Instead of spending 20 minutes polishing a prompt, the expert writes a lazy one and adds four words: use AskUserQuestion. The model then interviews you.
You click through its questions, and it keeps asking until, in the creator's words, it's impossible to fail. Why does this work? The model knows what information it's missing.
You often don't. I tested it on a messy brief and the model surfaced three questions I'd forgotten, namely deadline, audience and tone. That's the hidden cost of the perfect prompt.
You only pack in what you already thought of.
The same move saves your about-me file. Turn it into a Skill, and one command loads who you are, only when you call it. Zero maintenance, no clutter in every chat.
Talk, connect, and let the tool remember
Here's a bit of math from the post: you type around 60 words a minute, but you speak about 150. The fix is voice dictation. Hold shift, talk, and let a tool like Wispr Flow capture it.
The real win isn't speed. It's that you ramble when you talk, adding background, edge cases, context you'd never bother typing. More talking means more context means better output.
Then there's the morning inbox. The contributor wires Connectors into Gmail, Slack and an AI notes app, then asks for one summary of what needs a reply and what's slipping. One prompt replaces the whole manual scan.
The inbox swap is the one I'd protect most. Attention in the first hour sets the tone for the day, and handing that triage to a single prompt keeps your focus for the real work.
Want to get the most out of ChatGPT?
ChatGPT is a superpower if you know how to use it correctly.
Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.
Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.
*Ad
Never build the same thing twice
The last habit might be the most valuable long term. The second time you run a task, you tell the model three words: make it a skill. Your one-off workflow becomes a reusable tool.
Same logic kills spreadsheet busywork. Drop a messy CSV in, describe the output in plain English like monthly revenue by product with a growth column, and the formulas get written for you. One click sends it to Google Sheets.
When you have no idea where to start, the creator adds one more trick: a planning command that maps a task from A to Z before you touch it. I lean on this for anything with moving parts. It turns a blank page into a checklist you can work straight down.
Reading seven fixes is the easy part. Turning them into reflexes takes reps, and 3 Minute AI gives you those in three-minute lessons that each end with a task you run in a built-in chat lab, with a claude track built for exactly this. Daily lessons and two full courses are free.
The pattern behind all seven is one idea: stop doing manual setup, and let the tool ask, listen, connect and remember instead. That's the whole shift. If you want a starting order, add use AskUserQuestion to your next lazy prompt, then give voice a full day.
Your ten-minute cleanup
Open Claude right now, delete the junk folders feeding it stale context, and leave yourself just Skills and Projects.
The next time you catch yourself polishing a prompt, write a lazy one, tack on use AskUserQuestion, and let it interview you, and the same move for turning repeat tasks into skills is laid out in the exact fixes for all seven habits.
Worth 10 minutes if you keep redoing the same Claude work and want it to stick as a reusable tool instead.
Five reads to rebuild your Claude setup
Anthropic's guide to Skills explains how they load only when a task needs them, which is exactly why they beat pasting an about-me doc into every chat.
Anthropic on Projects covers how a Project holds the context for one piece of work, the other half of the two-part setup.
The Connectors getting-started page walks through Settings, then Connectors, to wire Gmail and Slack into Claude for that morning inbox summary.
VelvetShark's "let it interview you" walkthrough shows the AskUserQuestion pattern in practice, plus a reusable
/interviewcommand.Unite.AI's Wispr Flow review tests the talk-instead-of-type claim and notes where the speed gap shrinks.
Turn today's calls into tomorrow's proposal
Wispr Flow Notetaker brings your meetings into Claude or ChatGPT through a built-in connector. Ask for a proposal draft or a deal recap, and your AI works from an accurate record of what was said.
*Ad
Poll: You get one Claude habit to fix this week. Which one?
Hit reply and tell us why.
