8 Comments
User's avatar
Mike Romano's avatar

I’d say prompts still steer, but externalised context (files, folders, frameworks) now builds the stage. We’re moving from prompting into orchestrating. Prompt, context, and stored memory (like personalised settings or prior threads) all play their part…and the balance shifts dynamically.

Expand full comment
Alex McFarland's avatar

Perfectly said

Expand full comment
Omer Feinholtz's avatar

Very thought provoking! I think it's more of a 50/50 equation between prompt and context engineering - one without the other will give you a generic and mediocre result

Expand full comment
Christine Olukere's avatar

Exactly what I needed to read today! Thank you kindly Alex :)

Expand full comment
Digital-Mark's avatar

I just realised that I was a context engineer before I read your article. As I was using AI for cybersecurity with context and indicators. Keep up the good work. 😁

Expand full comment
The Market Maverick's avatar

The concept of AI Context Libraries is spot on. I’ve been exploring how to apply something similar to track the USD’s long-term weakening trajectory by layering together less-obvious indicators: real-yield differentials, reserve composition changes, geopolitical alignments, etc. I outline some of this thinking in my recent piece:

https://themarketmaverick.substack.com/p/the-dollar-is-quietly-declining

Anyone else experimenting with AI tools to contextualize macroeconomic signals over time?

Expand full comment
COUNT FROGULA's avatar

the designer should be designing a designer designing design by design if the designer in this case lacks design experience. you know what i mean? 🪰🐸🪰

Expand full comment
Casandra Campbell's avatar

Spot on! Context dramatically improves output and files and instructions help to scale it!

Expand full comment