The multi-model approach can be adapted into one’s life in a variety of ways.

But what about general prompting and other common uses of chatbots and AI tools?

This is the framework you need to know for how to pick which models play which roles and when and how to switch to a different model in the process.

The Foundation

The first model you would want to choose is the foundation model. For me, since I really like Claude and it is one of the best frontier models, Claude is my foundation model.

Therefore, once I have a new prompt or something new I want to research or chat with AI about, Claude is my first stop. And it can be any model within Claude. You don’t have to get that specific about which model you may use as your foundation model.

I highly recommend picking a model that you find yourself liking a lot, even if you are willing to pay. If someone asked me, “Which AI model have you tried that has had the best results for you?”, I would answer Claude by a million miles.

But it’s often not the same for everyone. One of my friends prefers ChatGPT. My mother prefers Google Gemini as her foundation model.

If you are not sure, peek your nose around the model world and try different ones to see which is the best fit for you overall.

The Stress Tester

The next model is the one that double-checks and critiques the foundation model.

This is probably as important as, if not more important than, the foundation model, as it will often have the final say in the matter.

After prompting your foundation model for a little bit and getting the answers you want, that means it’s time to move to the stress tester model.

This model should be:

1. Another model you may have had a good experience with while poking your nose around.
2. A model that maybe your company or school pays for.
3. A runner-up to the foundation model.

The main job of this model is to really push and challenge the foundation model’s stance or output. It is to act almost as a guide toward your goal. Almost like a second draft of what you are looking to do.

This is important because you want to take advantage of multiple brains that may help you flesh out an idea better. Think back to what I said about how this is like a first draft. Often, you don’t just publish your first draft right off the bat, or at least you shouldn’t. Usually, you have a trusted peer or adult read over it as an outsider to what you wrote and then give you feedback so that it is more polished.

The same kind of model applies here.

Now you might ask: “Evan, why not just use the foundation model again to polish it up or make a new chat to polish it using the foundation model?”

My answer is that doing that would defeat the purpose. Doing so is just like having the same person review it twice and try to come up with more feedback on something they already saw.

This is not nearly as effective. The stress tester is a crucial step, so please ensure that you choose a different brain to do the job.

The Integrator

You have finished the work and polished it up with the stress tester, maybe edited it yourself, and now feel good about it all around.

But execution is key to many things. You now need to do the thing with your product: make the decision, call the shot, create the piece, code the app, write the article, etc.

This may not always be a step you have in your process, as it is highly dependent on what you were trying to do from the start and whether that end goal changed at all along the way.

The integrator is often the AI tool that helps you with the pure execution of the idea.

For example, for this article, I wrote all the ideas, but then I went into Microsoft Word and used the built-in Microsoft Copilot tool to help me integrate this into an article for Substack, LinkedIn, and my personal website while helping with formatting and polishing up any hanging details.

For decisions, it might be Copilot or Claude built into Word helping you write a document to your superiors or those under you at your job. For an email, it might be using the Gemini built-in tool to help draft the important email that calls a meaningful shot. It all depends.

The fundamental idea is that the tool you choose here is made to operationalize the idea. Putting words into real actions and even executing those actions.

An Important Note

Also note that depending on your end goal, not every task needs all three models.

A quick email may only need a foundation model. A career decision or business proposal may deserve all three stages.

Final Thoughts

This may seem like a lot for a single task. But if you use this kind of framework for some of your larger or more impactful tasks, you have the ability to avoid representing only one brain in a decision or piece.

This may also help with costs, as dividing the work across models, some free and maybe even some paid, may be cheaper than having one powerful model do everything for you and eat up all your credits.

Choose your tools wisely and ensure your voice is sustained throughout the process!