What is the AI setup for an experienced dev starting on a new project?

Question

I have been working as a software developer for more than 20 years now. Been around the block a few times. Thanks to generous allowances in my current place, I've dipped my toes into LLM world and am quite impressed with it. However, we are on a legacy code base, LLM impact is limited in a sense. If I was to start a new project (let's say a web app with backend) - what is the state of the art for the setup? From Issue tracking, CI, automated deployments, agents, documentation etc - is there some blueprint I can just deploy and know that I am standing on the shoulders of giants?

joshuaayson · Accepted Answer

Use agent mode in vscode with sonnet or Claude code terminals and do an architectural decision record doc for each setup areas and all technical decisions. Ask more of your agent about your processes. Use context to your advantage and think not just of the product but also the docs as being the foundational context for the work.

edwmurph · Answer

One thing that's worked well for me is scattering agents.md files throughout the repo and loading the relevant ones based on what the agent is touching. It keeps architecture notes and local conventions close to the code, which works better than one giant project document.

bohdanstefaniuk · Answer

My approach to work on new project- Setup proper CLAUDE.md or AGENTS.md- Use cmux or conductor to manage multiple instances of Claude CodeWhen working of some business feature I always ask AI to summarize every conversation in following format: problem, brief solution, where change were made. And I store it in the repo under "journal" folder where each file has "2026-06-08-add-offile-firts-support.md"Also for me personally having separate repository with some boilerplate code become a huge booster of productivity.Usually inside this boilerplate repo I have:- Descriptions of desired project structure, so every project follows the same structure I personally like- Some CI/CD templates, in my case: build docker container, push into github registry, deploy to my bare bone VM- Authentication + third parties integration- Custom exceptions and middleware to handle those exceptions- And more