In a world where AI is everywhere, using just one model seems old-school. Here's how I'm trying to combine the best of multiple models for better results.
I read a great blog post on Anthropic's website about Building Effective (AI) Agents. One section, titled "Parallelization," describes running the same task multiple times to get diverse outputs. This approach mirrors what I've been doing across different foundational models like ChatGPT, Claude, and NotebookLM.
Embracing AI-First Workflows
My goal this year is to try to do all of my workflows with AI first. Even though it may be harder upfront (there's still a learning curve here), I think long-term it will pay off to get good at using all the systems to be more efficient in my work.
AI Multi-tasking
One thing I find myself doing A LOT, is combining use of different foundational models while working on a single project.
For example: I was recently producing some Brand & Product positioning messaging and I:
- Started with NotebookLM by uploading a bunch of background docs, youtube videos and web links with a blueprint for an output I was looking for - and it created a great first draft based on all the background I fed it.
- Refined with ChatGPT and Claude: I then took pieces of the document that I wanted to further workshop and I put them into both ChatGPT and Claude with the same prompt to get their feedback. I evaluated the feedback from each of the models, modified & combined parts from each, and produced a final product that I felt was way better than by just working in one model alone. Amazing.
Leveraging AI as Advisors
I'm one that loves to get feedback from other people on things I'm working on - I'll usually have my business partner read emails I'm going to send, work-product, proposals, etc. Now I also have 3 new AI advisors that I ask their opinion on things - and I'll often ask all of them the same thing to see if there is consensus :)
Enhancing Data Analysis with Multiple Models
We work with data a lot and working with data is another area where AI is becoming more useful. I'll often throw a spreadsheet into both Claude and ChatGPT with the same analytics prompt. With Claude’s experimental analysis feature enabled, I’ve been impressed by its outputs. Comparing responses from both models will often provide little nuances for me to think about, allowing for more comprehensive analysis.
The Future of AI Collaboration
While this isn't exactly what Anthropic meant by "Parallelization", the principle is similar. In summary, Here's generally how I've been using each model in my workflows:
- NotebookLM: Context-heavy tasks requiring synthesis of multiple sources
- Claude: Analytical tasks and more technical writing
- ChatGPT: Creative ideation and brainstorming
As these models just keep getting better, I think the pace of using them together is just going to accelerate. Combining their unique capabilities has already helped me a lot.
P.S. AI isn’t taking my job just yet.
P.P.S. I asked both ChatGPT and Claude if this post was strong and they kind of said "meh" - but what the hell do they know.