Thank you for your comment ! :-) So many improvements are suggested every day, that I spent more time at the moment improving my coding agents instead of focusing on the Agentic AI Systems I develop. Hope to be able to share more about Agentic AI in general after my holidays!
This is really helpful🫡 Funny thing is i just created an agents. Md on my personal project because i got tired of repeating myself , and i still had to repeat myself on certain occasions. Now i have a new challenge
The cold memory layer really caught my attention, the idea of agents fetching spec docs on demand instead of cramming everything into one file makes so much sense. I'm completely new to retrieval systems so it feels like the next frontier for me to explore. How would you adapt that 3-layer setup for a smaller solo project? Does the overhead make sense at that scale, or is there a lighter version of the same idea?
Hi, for a medium-sized code base, I would always maintain a hot memory layer that is continuously loaded. You probably don’t need 19 agents. A smaller setup is often sufficient: for example, one agent for planning and research, one for architecture, and one developer agent. Depending on the project, you might also add a UX-focused agent.
For the cold memory layer, I’m currently experimenting with a small number of specifications that are well separated by topic and kept relatively concise. I think we are all learning at moment, what works best. Would be great to hear, what your strategy is and what you experienced ?
Hi Ursula, thanks for sharing - not my world but I can see this would be incredibly useful for anyone working with AI coding tools at scale. 🙌🏼
Thank you for your comment ! :-) So many improvements are suggested every day, that I spent more time at the moment improving my coding agents instead of focusing on the Agentic AI Systems I develop. Hope to be able to share more about Agentic AI in general after my holidays!
This is really helpful🫡 Funny thing is i just created an agents. Md on my personal project because i got tired of repeating myself , and i still had to repeat myself on certain occasions. Now i have a new challenge
Great to hear, that my post was helpful for you ! What Agentic AI topics are you most interested in ?
The cold memory layer really caught my attention, the idea of agents fetching spec docs on demand instead of cramming everything into one file makes so much sense. I'm completely new to retrieval systems so it feels like the next frontier for me to explore. How would you adapt that 3-layer setup for a smaller solo project? Does the overhead make sense at that scale, or is there a lighter version of the same idea?
Hi, for a medium-sized code base, I would always maintain a hot memory layer that is continuously loaded. You probably don’t need 19 agents. A smaller setup is often sufficient: for example, one agent for planning and research, one for architecture, and one developer agent. Depending on the project, you might also add a UX-focused agent.
For the cold memory layer, I’m currently experimenting with a small number of specifications that are well separated by topic and kept relatively concise. I think we are all learning at moment, what works best. Would be great to hear, what your strategy is and what you experienced ?