Solving tricky problems...

GenAI success reaquires expertise + space + creativity + trust

Duncan Anderson
2025-01-15

There's two things I always say to people:

  1. Don't lock yourself into a single vendor - the space is moving fast, vendors jump over each other frequently and different models have different pros/cons.
  2. Be open-minded - adopt an approach that allows for creativity and the ability to explore new avenues. Give your team the tools and freedom to "play" – they will likely come up with new ideas.

Here's a good (and real) example of a conversation from today:

Me: That's really cool, how did you do it?

Engineer: Well, I chatted with Claude to develop a meta-prompt, sent that meta-prompt to Gemini to create a description of an image, then used that description to generate an image using Runware.

That's three different models, from three different vendors, all collaborating to solve a tricky problem (that all the image generation models were incapable of generating safe representations of what we were asking).

We gave our engineer a broad objective and let him tinker with different ideas to solve the problem. We knew the problem could be solved, we knew our engineer was resourceful and good at "figuring stuff out". He needed the space to play with stuff and see what he could get working, so that's what he got. And he cracked a complex problem.

I find this a lot in GenAI - space, freedom, creativity, trust are as important as expertise.