I didn’t switch to Claude.

I promoted it!

Like you, I started by using ChatGPT. Then Gemini came out.

It was better. Especially at Deep Research. So I switched.

When I put together my AI Equity Analyst framework detailing how I built AI into my investing process, it was based on Gemini w/ Deep Research.

Then Claude leapfrogged Gemini.

I was tempted to switch again, this time to Claude. But I figured out a better way.

I promoted Claude to ‘Senior LLM.’ 🙂

Now, my main workflow is built around Claude. However, while it’s the best model overall, it’s not the best at everything.

For example, Gemini 3 Nano Banana is still the best at image rendering. Grok has native integration with X and can search it better than Claude can, etc.

So when a task comes up where another LLM is the best solution, I ask Claude to come up with the high-level answer and then create a prompt geared towards the other LLM to execute the details. 

(For example, Claude zero-shot the org-chart image in this post by creating a detailed prompt for Gemini).

AI will keep evolving. A lot.

The important thing is to systematically build it into your process and continue to iterate. That’s the only way you will stay on top of it and improve your process.

Compound With AI - what do you think of this approach? By the way, I love your work on AI in investing, great job!