2 minute read

TL;DR:
I have been quiet since Feb 20 because I was heads down finishing the AssetTrack MVP. As of March 15, the full custody lifecycle works end-to-end: issue, preview, commit, return, and slot reconciliation. The system now runs cleanly in Docker with SQLite and passes a full operator smoke test.

Project page: AssetTrack

Context

If you follow this blog you probably noticed something unusual:
no AssetTrack updates since February 20.

That was not because the project stalled.

It was the opposite.

I went heads down to finish the operational seams of the system so the MVP could actually work end-to-end without operator dead ends.

AssetTrack is an offline-first asset custody system designed for field environments where laptops are issued and returned from Pelican cases with fixed slots.

The system tracks custody through an append-only event log and projects the current state from those events.

But architecture only matters if the workflow works for the operator.

So the last month was spent closing the final seams.

What Changed

Between Feb 20 and March 15 the following milestones landed:

Workflow Stability

The entire operator lifecycle is now validated:

login
select holder
scan assets
preview
commit
verify custody
return
verify case occupancy
verify dashboard

The lifecycle now works cleanly:

STORAGE → IN_CUSTODY → STORAGE

Key Fixes

Several small but critical workflow issues were fixed:

  • Holder context now persists after issuing assets
  • Queue timestamps display in UTC
  • Return confirmation shows the final slot location
  • Session activity refresh works reliably
  • Preview pages now clearly signal readiness to commit

These changes eliminate the subtle workflow dead ends that appear in real operator use.

Operator UX Improvements

Technical language was replaced with operator language.

Preview pages now read:

Issue preview

Ready to Issue
Review the listed assets below. Commit Issue is the next step.

Return preview

Ready to Return
Review the listed assets below. Commit Return is the next step.

This removed internal terms like “Per-Asset Diff”, which made sense to developers but not to operators.

What I Learned

The hardest part of systems like this is not the data model.

It is the workflow seams.

Everything looks correct in isolation:

  • database schema
  • event model
  • projection logic

But the real test is whether a human can run the workflow repeatedly without confusion.

That is why the last few weeks focused almost entirely on operator smoke testing.

Every issue fixed during this period came from running the workflow like an operator would.

Next

With the MVP complete the project moves into the next phase:

  • operator workflow polish
  • navigation simplification
  • deployment packaging
  • documentation improvements
  • real field testing

The important milestone is this:

AssetTrack is now operational.

It works end-to-end.

And that is the foundation everything else builds on.

Updated: