✅ Org Block End-to-End (Almost)
2 min read

✅ Org Block End-to-End (Almost)

✅ Org Block End-to-End (Almost)
Photo by Tandem X Visuals / Unsplash

It’s 10:50pm on August 6th, and I’m calling it. Jiu-jitsu in the morning.

Tonight I finished the full layer execution path for the Organizations block, both backend and frontend. This was the first time I explicitly structured the work as blocks and sub-blocks, not just loosely scoped features. That mental shift actually held up in practice. The terminology feels solid. I know where things go, how pieces relate, and how to run the plans cleanly. That's good.

What’s working:

  • Ran both the backend and frontend plans start to finish, including all the prompt prep and hardening.
  • Executed both sets fully (as the anvil, manually).
  • Got all tests passing.
  • Lint and structure audits are clean.
  • The org models, services, and UI components are all in place.

What’s not clean yet:

  • Session management — there are backend bugs around checkSession, and the routing structure still needs alignment across repos.
  • Testing the backend in isolation — since I’m working with split codebases now (unlike Simrata where I had both codebases in one Cursor window), I don’t have the same delegated flow. I’ve got to manually hit backend routes using Postman, but I haven’t yet traced where to pull the right auth tokens from the frontend.
  • Expected minor mismatches between frontend/backend infra still need ironing out.

The upshot: I might already be done with the whole org block. I just don’t know yet. Tomorrow I’ll find out if the wiring actually lets me access and interact with the thing I’ve built.

But from a dev muscle perspective, this was a solid rep. First end-to-end execution under the new block-based system, and it held up. The new plan scoring rubric is causing the bulk-generated layer prompts to pass their (new, improved, harder) scoring hurdles on the first try too, saving a ton of time. There are usually 5-10 layers, so not having to go back and forth with the LLM to fix the individual layer prompts is already quite encouraging.


What’s next: Test backend APIs directly. Confirm session auth. Verify the frontend can actually hit the org endpoints cleanly. If that works, this block is done.