I Spent Six Months Fighting With Claude

You open Claude for the third time today.

Same project. Same goals. Same frustration.

And once again, you're typing out the context from scratch:

"I'm working on a content strategy for our B2B SaaS company.

Our audience is mid-level managers who are overwhelmed by digital transformation but sceptical of new tools.

We've tried educational content but engagement is low. Leadership wants more thought leadership but I'm not convinced that's the answer.

Our brand voice is professional but approachable, data-driven but not academic..."

Ten minutes later, Claude gives you advice that could work for literally any company in any industry.

Generic frameworks. Surface-level suggestions. Nothing that acknowledges you've been wrestling with this exact challenge for three weeks.

You close the tab. Again.

Here's what nobody tells you about AI productivity:

The problem isn't Claude. The problem is that you're introducing yourself to it every single day, like it's the first time you've met.

I Was Doing The Exact Same Thing

Six months ago, I started using Claude to help grow my newsletter.

I'd upgraded to the paid plan. I'd read the productivity guides. I'd even created a "Project" and uploaded my best-performing posts as reference material.

It didn't matter.

Every conversation still started from zero:

• "What's my audience?"

• "What's my content style?"

• "What topics have I already covered?"

• "What actually gets engagement?"

By the time Claude had enough context to be genuinely helpful, I'd spent 20 minutes just preparing it to assist me.

I was doing this 4-5 times per week.

That's over an hour every week spent teaching AI the same lessons it forgot overnight.

The promise was that AI would save me hours. Instead, I was spending those hours in what I now call the "Context Loop From Hell."

And I know you're stuck in it too.

The Moment Everything Changed

Three months ago, I was drafting my weekly newsletter when I had a realisation that felt embarrassingly obvious:

I wasn't treating Claude like an AI assistant. I was treating it like a search engine.

Think about how you work with a real colleague:

You don't re-explain your company's mission every morning. You don't describe your audience from scratch before every project meeting. You don't hand them last quarter's performance data and say "by the way, this is what we learned."

You onboard them once. Then they remember. They get smarter about your work over time. They start making suggestions that actually reflect your specific context, not generic best practices.

That's what I'd been missing with Claude.

I wasn't giving it a proper onboarding. I was giving it a rushed briefing before every single task.

So I stopped.

I Built What I Call a "Persistent Intelligence System"

Instead of treating Claude like a blank slate every day, I built it a brain and gave it a memory.

The Brain: A master prompt that defines who I am professionally, what I'm trying to achieve, how I think, and how I want Claude to approach every conversation about my work.

The Memory: A knowledge base containing my past content, performance data, reader feedback, what's worked and what hasn't, my strategic frameworks, and my voice guidelines.

Now when I ask Claude about content strategy, it doesn't pull from generic marketing knowledge.

It references my specific audience data. It analyses patterns from my highest-performing posts. It suggests approaches that align with my unique voice and positioning—not someone else's.

The transformation was immediate.

Here's What Actually Changed

Before the system:

Me: "Help me come up with newsletter topics for this week."

Claude: [Generates 10 generic AI/productivity topic ideas that could work for any newsletter in my space]

After the system:

Me: "Based on my content performance data and recent reader comments, what newsletter topics would drive both authority positioning and subscriber growth?"

Claude: "Looking at your Q3 performance, your audience responds strongest to transformation stories (avg. 47% open rate) versus tactical tutorials (31% open rate). Your 45-55 age bracket—your fastest-growing segment—engages 3x more with career-transition narratives around AI adoption anxiety.

Your highest-performing piece was 'The 51-Year-Old Analyst Who Felt Obsolete'—it drove 340 new subscribers, 67% above your baseline.

I'd suggest building on this pattern: Find another transformation story, but focus on a different industry vertical you haven't covered yet. Your readers in healthcare and education have been underrepresented in your case studies, but they represent 23% of your subscriber base."

That's not generic AI advice. That's a strategic partner who actually knows my work.

In four months since building this system:

• My newsletter grew from 0 to 4,000 subscribers

• My content production increased 47% without working longer hours

• My engagement rates improved because every piece was informed by actual performance data, not guesswork

But here's what matters more than the numbers:

I stopped fighting with Claude. And started collaborating with it.

I'm not re-explaining context anymore. I'm not getting generic advice anymore. I'm not wondering if Claude actually read the documents I uploaded anymore.

Every conversation builds on the last one. Every suggestion gets smarter. Every interaction feels like working with someone who actually gets what I'm trying to build.

You're Probably Wondering: "How Complicated Is This To Set Up?"

That was my first question, too, when I started seeing people talk about "custom AI systems."

It sounded technical. Like something that required coding or complex integrations or hours of setup time, I didn't have.

It's not.

Building a persistent intelligence system takes about 2 hours of focused work. Once.

Then it just works. And it keeps getting better as you add to it.

Here's the simple truth: You're already spending more than 2 hours per month re-explaining yourself to Claude. You're already uploading documents that don't seem to help. You're already frustrated that AI productivity tools aren't making you more productive.

This is just redirecting that wasted energy into something that actually compounds.

The Two Components That Make This Work

Your system has two parts working as one unified intelligence:

Part 1: The Master Prompt (The Brain)

This is a structured document that tells Claude:

• Who you are professionally and what you're trying to achieve

• Your communication style and preferences

• How you want Claude to think about different types of requests

• What "good work" looks like in your domain

• The frameworks and principles that guide your decisions

This stays consistent. It's Claude's operating system for understanding your work.

Part 2: The Knowledge Base (The Memory)

This contains:

• Your past work and what you learned from it

• Performance data and patterns you've identified

• Strategic documents and frameworks you reference

• Examples of your voice and style

• Feedback from clients, readers, or stakeholders

This grows over time. It's Claude's context for making intelligent suggestions.

Together, they transform Claude from a generic assistant into something that actually knows your business.

Here's What This Looks Like In Practice

Let's say you're a marketing manager at a SaaS company.

Without the system:

Every time you need campaign ideas, you're explaining:

• Your target audience demographics

• Your positioning in the market

• Past campaigns and their results

• Budget constraints

• Timeline pressures

• Brand voice guidelines

And Claude gives you ideas that sound good but don't reflect what's actually worked for your specific audience.

With the system:

You ask: "What campaign angle would work best for our Q4 push?"

Claude responds: "Based on your Q2 and Q3 data, your audience (operations managers at 50-200 person companies) responds 3.2x better to efficiency/time-savings messaging versus cost-savings messaging, even though they cite budget as their primary concern in surveys.

Your 'automated workflow' campaign in Q2 achieved 18% conversion—your best performer this year—because it focused on 'getting home by 6pm' rather than 'reducing operational costs by 30%.'

For Q4, I'd suggest a campaign built around 'Reclaim your strategic time' positioning, featuring real testimonials from ops managers who went from firefighting to strategic planning. Use your successful Q2 visual style—simple interfaces, no stock photos, real Slack screenshots showing the before/after."

That's the difference between generic AI and AI that actually knows your work.

Why This Matters More Than You Think

We're entering a weird moment in work history.

75% of knowledge workers now use AI tools daily. But only 12% feel like they've actually integrated AI into their workflow successfully.

That gap is creating a new type of inequality.

On one side: People who've figured out how to make AI genuinely useful. They're becoming more productive, more creative, more strategic. They're getting promoted. They're delivering results that seemed impossible a year ago.

On the other side: People stuck in the Context Loop. They're using AI, but it's not helping. They're spending time on AI that they could've spent actually doing the work. They're falling behind without understanding why.

The difference isn't talent. It's set up.

The high performers aren't working with better AI. They're working with AI that's been properly onboarded to understand their specific work.

What The Next Two Hours Need To Look Like

You can keep doing what you're doing:

• Opening Claude

• Typing out context

• Getting generic advice

• Closing Claude frustrated

• Repeat tomorrow

Or you can spend two focused hours building a system that ends this cycle permanently.

Here's what those two hours look like:

Hour 1: Build Your Master Prompt

• Define who you are professionally (15 min)

• Outline what you're trying to achieve (15 min)

• Describe your communication preferences and style (15 min)

• Identify the frameworks that guide your work (15 min)

Hour 2: Create Your Knowledge Base

• Gather your 5-10 best pieces of past work (20 min)

• Document what worked and what didn't (20 min)

• Add any strategic documents or guidelines you reference regularly (10 min)

• Write brief context notes for each piece (10 min)

That's it.

Two hours of setup that compounds every single day after.

I expected the system to save me time on repetitive tasks.

It did that. But it did something more valuable:

It made me better at my own work.

Because Claude now remembers everything I've learned, I'm forced to actually document what works and why. I can't just move from project to project, hoping I remember the lessons.

The system became a forcing function for capturing knowledge I would've otherwise lost.

When I analyse why a newsletter performed well, I add those insights to the knowledge base. When I discover a new content angle that resonates, I document it. When a reader gives feedback that shifts my thinking, I save it.

The system gets smarter. Which means I get smarter.

It's not just an AI tool anymore. It's a second brain that remembers everything I've learned about my work—and helps me build on it.

The Choice That Defines The Next Five Years

Here's what's actually happening in knowledge work right now:

One group of people is building AI systems that compound their expertise daily.

Another group is starting from scratch every time they open a chat window.

In six months, the productivity gap between these groups will be undeniable.

In a year, it'll determine who gets promoted and who gets passed over.

In five years, it'll determine who thrives in knowledge work and who burns out trying to keep up.

This isn't about being "good with AI." It's about understanding that AI without memory is just an expensive search engine.

AI with memory is a colleague who never forgets.

You've already spent months fighting with Claude. You've already invested time and money in AI tools that haven't delivered on their promise.

Two hours. That's what stands between the frustration you feel now and the transformation you've been chasing.

The people who figure this out first won't just be more productive. They'll be operating in a completely different league.

But here's what you can do right now:

Open Claude. Start a new project. And ask yourself one question:

"If I had a human colleague joining my team tomorrow, what would they need to know to actually be useful on day one?"

Write that down. All of it.

That's your master prompt. That's your brain.

Everything else is just adding memory to it.

New to AI but curious about what's possible?

Subscribe here for weekly tutorials that actually make sense.

No jargon, no hype, just step-by-step guides you can follow.