Abdullah Akbar Khalid
← All exhibits

Exhibit E · Operations

Automations need somewhere to land, and somewhere you can see them fail

One dashboard for what six apps and my head used to hold, and one page that tells me whether the machinery behind it is still running.

Status
Public
Stack
React · TypeScript · Vite · Supabase
Source
github.com/akbar-33/command-center-companion

The problem

I run a lot of my working life through AI automations: briefings, task capture, inbox triage, follow-ups. Those automations need somewhere to put their output, and I need somewhere to go to see whether they are still working.

Nine surfaces, one honest readHover any surface
Today
Tasks
Overdue
Calendar
People
Jobs
Score
Remaining
Automation Health

The page that justifies the project

Automation Health. Automations fail silently, and that is the entire danger. A briefing that stops arriving looks exactly like a quiet news day until you go looking, by which point you have been quietly uninformed for a fortnight.

The page shows whether each background job actually ran, when it last succeeded, and what it expected its own schedule to be. It is the difference between a dashboard that reports on the world and one that also reports on itself.

Decisions worth defending

Row-level auth from the start

Personal data, contacts and application pipeline live here. Access control belongs in the database, enforced per row, rather than in the client where it is one bug away from not existing.

Capture from automated sources, not just by hand

A task list only I can write into becomes stale the first busy week. Edge functions let the automations file directly into the same backlog I read.

Back to all exhibits