It's Alive

It’s been a while since my last update, but ResumoFII is finally live.

This week I shipped it in it's currently half-baked state and it's available at resumofii.com.

I was starting to get tired of being stuck in an endless build loop without validating what I have so far in the real world.

I subscribed to Claude Pro and started pushing features out the door. Shipping something real is always better than planning the perfect system. It's been a game changer and I should've adopted this way earlier in my workflow.

Getting a basic CI/CD setup with deployment on Vultr took only a couple of hours of iterating, testing docker compose files and github actions with secrets but it's deployed now. This will be the driver to move faster and soft-launch in the coming weeks.

Adding new features with Claude Code is a cinch since I already had a sensible architecture in place, so it was just a matter of generating some markdown files to describe how I want new features to be built.

I settled on the following setup:

• CLAUDE.md
  • Describes the ResumoFII monorepo starting with the directory structure of the repo itself and available npm and cargo commands.
  • Describes the frontend SvelteKit architecture including API communication, file structure, authentication flow, key data types and error handling.
  • Describes the backend Rust architecture including the clean architecture layers, implementation order for new features, code quality standards, the authentication pattern, the repository and database patterns and finally the testing strategy.
  • Describes the development setup and environment variables.
  • Gives an overview of the API.
  • Describes naming conventions.
  • Describes production deployment using docker and the CI/CD pipeline.
• README.md inside docs/agent-instructions
  • Describes that there are step-by-step implemention guidelines available in markdown format (focused on the backend for now) and guides Claude to utilize feature-template.md when starting new work.
• feature-template.md inside docs/templates
  • An in-depth overview of all the steps needed to implement a new feature. This is a detailed explanation of all layers that need to be modified and in which order. Gives guidelines on how to test and break down the feature in small workable steps.

This workflow has been pretty good so far and keeps Claude from going off the rails. I've built an admin level batch configuration feature this way and I'd say it was implemented reasonably fast with me mostly confirming edits via Zed's terminal.

In production I have the ability to toggle the summarization on or off per fund or per batch. So I've enabled it so far on 4 FIIs to check Gemini's output but so far I've been happy with the results.

I need to monitor the cost of running this project, so I'll scale up batch sizes carefully to keep costs under control.

Now that the app is actually live I can slowly increase the size of the batch and iterate on the prompt as it's not 100% compatible with all the different fund types an investor can find on B3.

After ingesting some more data I'll work more on distribution and soft-launch this on Reddit to see if anyone else finds it useful.

The core project is now live and I need to continue to iterate and hopefully draw the attention of some potential users.

Now that it's live, time to find out what breaks.