And once folks have nailed this, it'll be time to set these much improved prompts inside a project or gem in order to give you advice on specific repeatable topics.
“This is what researchers call the “Foundation Problem.” The AI’s foundation is generic training data. Without your specific context added to that foundation, you’re stuck at 10% results.”
Double down on this I see it too often. Great article thanks!
Great point, Joel. Context for the prompt is everything. One of my tricks is to use the voice dictation feature and give it as much 'why' and 'how' context for prompts, otherwise, I tend to just write a handful of words like Google search.
If you don't brief the AI with the same rigor you'd use for a senior analyst or a consultant, you'll be limited to generic output. The idea that a smaller, older model with rich context outperforms a massive, new one with none is a profound observation.
I recently read a study on that last point. So many are promoting new models as if they were Google, and getting no better results than the model 2 yrs ago
Do you use Claude Skills, Joel? I'm a huge fan of using them for priming the conversation before you even begin the work, so many possible applications!
I really don't, I only use projects because I keep tweaking them and that's not easy to do with skills, but I actually need to spend more time with skills
And once folks have nailed this, it'll be time to set these much improved prompts inside a project or gem in order to give you advice on specific repeatable topics.
Great writeup!
That’s exactly it! Preload the context as much as possible and get that kind of granularity each time. You get it, Malcolm!
“This is what researchers call the “Foundation Problem.” The AI’s foundation is generic training data. Without your specific context added to that foundation, you’re stuck at 10% results.”
Double down on this I see it too often. Great article thanks!
So well said! These type of insights are why I will reach out soon for a collab or guest post in Jan if you are interested in writing one
It would be my honor!
I’d love to do a collab or guest post. Please reach out when available, and I’ll set a reminder to do the same in Jan. 🔥
Sounds good! Loving your stuff
Treating AI like a consultant instead of a search engine is the simplest, most overlooked shift for leaders.
I really think so!
Great point, Joel. Context for the prompt is everything. One of my tricks is to use the voice dictation feature and give it as much 'why' and 'how' context for prompts, otherwise, I tend to just write a handful of words like Google search.
Oh I am become so reliant on voice dictation haha! It changes the game
This is an excellent point.
If you don't brief the AI with the same rigor you'd use for a senior analyst or a consultant, you'll be limited to generic output. The idea that a smaller, older model with rich context outperforms a massive, new one with none is a profound observation.
I recently read a study on that last point. So many are promoting new models as if they were Google, and getting no better results than the model 2 yrs ago
Love the prompt comparison. This is one of the key skills to build up an unfair advantage towards other professionals in your business.
Thanks, Florian!
This is a game-changer. Showing how a bit of context unlocks AI’s real potential is incredibly valuable.
Do you use Claude Skills, Joel? I'm a huge fan of using them for priming the conversation before you even begin the work, so many possible applications!
I really don't, I only use projects because I keep tweaking them and that's not easy to do with skills, but I actually need to spend more time with skills
you?
Obsessed with them. Definitely reckon you'd make them sing!
I'm convinced :) setting some time aside for that after thanksgiving
Great prompts, and very interesting how change the capabilities just by adding the context.
Yes! If you try it out, let me know how it goes :)
These prompts are awesome. So important to become comfortable using Ai to its full capabilities… or at least more than 10%
Glad to hear that!