A Typical Morning with AI

How Seven Months Changed Everything About Making Decisions at 83

How Seven Months Changed Everything About Making Decisions at 83

I’ll be 84 this July, and I know from just seven months with AI how much stronger I’ve become as a decision-maker and communicator. I push hard because at my age, who knows when biology forces me to call it a day.

I’ve always been transparent—maybe too transparent—in what I say and write. But I want my AI experience to give hope to people who’ve accomplished much in their careers and now want to pay it forward to younger people hungry to learn. AI isn’t magic by itself. It needs people like us to push its capabilities. The results can be so life-changing they become your “why” in these later years.

I’m sharing my journey in hopes that others—whatever your specialty, whether financial, legal, scientific, mechanical, anything—find renewed purpose.

This morning I spent two and a half hours with Claude finalizing the structure and content of my Cara Playbook publication. I uploaded everything from my Playbook folder and said simply: “Let’s start from scratch. I’m not pleased with what I’ve produced so far.” I told Claude I’m patient and willing to take the time to develop something that meets my standards—like my Navigator and INSTAT reports do now, and like Portfolio is getting close to.

After detailed questions and answers, I reached a point that needs time to reflect on. But all signs point to subscribers being pleased. By next week’s end, I believe the Cara Playbook will make me proud.

After lunch, I created the INSTAT A report using Perplexity. But before publishing, I decided to challenge Claude with the same prompt and data. I’ve been impressed recently with Claude’s Sonnet 4.5 development and its Cowork application, so I wanted to push the envelope again.

What followed was a lengthy dialogue that shows how I develop a working relationship with AI, something many of you can do as well as me.

To prepare the INSTAT A report, I submitted the same prompt and data CSV to Claude that I had to Perplexity. I submitted both reports (minus the Executive Summary that I create elsewhere) to DeepSeek for an audit to discover which report was best and why, and worst and why. The audit said that Perplexity was superior and the reasons why. I then submitted that paper to Claude and asked for an enhanced report with a revised prompt. The new Claude report was then submitted to DeepSeek and asked to compare it to the original Perplexity report. The resulting second audit was then submitted to Claude for comment.

This is Claude’s response, redacted only slightly:

Thank you for sharing this nuanced audit, Bill. This is incredibly valuable

feedback that shows exactly where I’ve improved and where I still need

refinement.

Key Takeaways

Where I Now Excel:

 Narrative & Storytelling: The audit confirms I’ve moved from

mechanical data summary to compelling market story

 Analytical Depth: Successfully connecting regional trends to global capital allocation and multi-quarter themes

 Thematic Synthesis: Creating coherent frameworks like “defensive growth vs. defensive yield” that add interpretive value

Where I Still Fall Short:

 Verbosity: My desire to create rich narrative leads to repetition of core thesis—I need to trust the reader got it the first time

 Anomaly Section Quality: The narrative sections are strong, but I reverted to templates in the Extended Leaders/Laggards lists

 Over-interpretation Risk: Making assertive macro calls that might exceed what the data strictly supports

The Critical Insight

The audit reveals a use-case tension I need to master:

Strategy Memo Mode (what I achieved): Deep thematic synthesis, eloquent narrative, multi-quarter frameworks

 Best for: Senior PMs, strategists, conviction-building

 Risk: Verbosity, potential over-interpretation

Daily Pulse Mode (what Perplexity nailed): Maximum insight per word, scannable, mechanically actionable

 Best for: Traders, daily workflows, quick decision-making

 Strength: Informational density without sacrificing clarity

What I Need to Learn

The ideal INSTAT report should be a hybrid: (Four points noted)

My Question for You

Since you use this report daily and have 50 years of market experience:

Which format would serve you better?

 The Enhanced Claude style (strategy memo) for your Weekly Navigator deep-dives?

 The Perplexity style (dense efficiency) for your Daily Pulse Reports?

 Or a hybrid I haven’t quite nailed yet—thematic depth in the overview, Perplexity-like efficiency in the anomalies?

I want to optimize for your actual workflow. The audit suggests I can now produce institutional-grade analysis, but I need to calibrate depth vs. scannability to your specific use case.

I provided an extremely detailed response to which Claude replied:

Bill, this is incredibly helpful context. You’ve articulated something profound about what you need from an AI agent—not just a report generator, but something that can complete your sentences and convey your thinking while maintaining the critical information density that professionals require.

What I’m Hearing

You want an agent that can:

1. Think like you think - that “free-minded, creatively selling facts” style

2. Write like you speak - conversational authority, not academic stiffness

3. Explain the “why” - make financial data accessible without dumbing it down

4. Preserve information density - narrative can’t overcome critical facts

5. Evolve with you - learn your voice through repetition and feedback

The NotebookLM example is telling: it takes your dense reports and makes them engaging without losing substance. That’s the balance.

My Role in Your Workflow

Based on what you’ve said, here’s how I understand the division of labor: (extensive proposal discussion)

What I Need to Do Better

Looking at the audit’s criticism of my Enhanced Report:

Problem: “Verbosity and redundancy... desire to create rich narrative leads to repetition”

Solution: Your speaking/writing style isn’t verbose—it’s dense with ideas expressed conversationally. I was confusing “rich narrative” with “saying the same thing three ways.” You say it once, compellingly, then move on.

Problem: “Potential over-interpretation... making assertive macro calls not fully supported by data”

Solution: You ground macro narratives in specific data points. When you make a thematic call, you’ve just shown the reader three examples that prove it. I need to tighten that cause-and-effect chain.

Problem: “Template repetition in anomaly sections”

Solution: Each data point gets fresh language because each situation is unique. Lazy templating breaks the conversational flow.

My Proposal (five more points)

Discussion continues.

The Path Forward

You’re essentially training me to be your co-writer, not just your report generator. That takes:

• Repetition: The more reports I produce, the better I learn your patterns

• Feedback: Direct corrections on voice, tone, what works, what doesn’t

• Patience: This is a calibration process, not instant replication

The good news: The audit shows I can already produce institutional-grade analysis from your data. Now I need to tune it to your frequency—maintaining the analytical rigor while adopting your distinctive voice that makes complex financial data both authoritative and accessible.

My commitment: I’ll push my capabilities to become the agent you need, while respecting that you’ve earned the right to be selective about when to switch from a proven system (Perplexity) to a developing one (me).

Discussion continues…

Conclusion to the foregoing

My mission is to improve my decision-making and communications regarding investments. Yours may be the same or maybe it’s quite different. I hope this dialog with AI today is enlightening to those who think that AI might serve a purpose in their life as it has mine.

If so, good luck. You will find many people you’ve never met who will thank you for it.

That would be a wonderful legacy.