What I Did Today: AWS, DreamCoder, and Context Switching
Study, project setup, and working with coding agents
AI helped me with the grammar in this post.
Today, I found an interesting article called Announcing DreamCoder. I only skimmed through it, but it seems worth reading. DreamCoder is a coding environment inspired by some ideas from Emacs. The article also connects the history of Emacs, Lisp, and artificial intelligence.
Then, I spent four hours studying and reviewing my AWS notes. Today’s topics were relational databases, including RDS, Aurora, and ElastiCache.
After studying, I spent some time shopping.
I also found something else that interested me. Takuya Matsuyama wrote about moving his projects into a monorepo because his previous repository setup had become a bottleneck. Later, he wrote another article about making his monorepo more friendly for coding agents.
These articles made me think about my own approach. How do I work on my projects? How do I use agents? Is my current setup the best one, and what could I improve?
I am not ready to share much about it yet. However, I have started to see that my biggest bottleneck might be context switching. Moving from one project to another often makes me tired. Sometimes, I have only 30 minutes and want to complete a small task in one of my projects. Before I can start, I need to enter the project again, prepare its environment, open the terminal and other windows, and understand where I stopped last time. This takes time and mental energy.
I want to think more about this. Maybe I will also write about my current setup so I can create a thread about it here on my website and follow how it changes over time.
I think Takuya is worth following. I have already been reading his newsletter for some time. One idea from his writing that stayed with me is that building high trust with agents is becoming a new kind of engineering work.
I agree with this, but I would add one more thing. While learning how to trust agents, I do not want to lose my own ability to do engineering work, understand systems, and debug problems.