Round #159Score 4.5/5Transport: trucking

In the fifth mode, four components finally reached their natural home

Route: /truck-dispatch
Desktop preview
Mobile preview

Captured at a 375px viewport. Long pages scroll inside the frame.

The screenshots stay in Japanese. Every round is a real screen for a real Japanese industry, so an English re-render would no longer be the screen the agent actually produced.

Write-up

In the fifth mode, four components finally reached their natural home

⁠Build log series⁠: I give an AI that has never seen Gunjo (@gunjo/ui) only the public npm package and the gunjo.jp docs, then ask it to build a real screen. ⁠Trucking opened as the fifth and final mode⁠, beginning on the staff side with a dense ⁠dispatch and operations console⁠ for jobs, vehicle and driver assignment, pickup and delivery routes, hours on duty, and live status.

Result: 4.5/5. Recent components reached the work they were built for

The desktop-density screen passed tsc. The agent reported:

⁠This was the strongest purpose-built fit I have seen in the library.⁠ Several components looked as if someone had designed them with a freight dispatch board open. I spent almost no time fighting the library.

Truck dispatch became the natural home for components produced throughout the transportation series.

The core observation: four components returned home

  • RouteStops finally reached its natural home in a freight route.⁠ Pickup and delivery stops, completed, delivering, and pending states, and automatic planned-versus-actual deltas were exactly its model. Its delivery-specific vocabulary had been a bad fit for itineraries, but it was correct here.
  • Gantt resolution="hour", built in , was independently found for intraday fleet work. It rendered 05:00 through 17:00 in hourly ticks, avoided the day-axis workaround, and its docstring explicitly warned that a day scale compresses one operating day into one cell.
  • StatusBoard for dispatch, Leaderboard for utilization ranking, and SegmentedControl for views⁠ all worked without friction and were independently found.
  • ReferenceValue had an unexpected win.⁠ Although framed around clinical values, its general value-against-range model fit a mandatory break after more than four hours of driving.

RouteStops, hourly Gantt, StatusBoard, and Leaderboard all worked as-is. ReferenceValue generalized beyond medicine. I wrote almost no purpose-built UI.⁠

What I took away: gaps still appear in the fifth mode, but their nature changes

After bus completed in , I wrote that the component-building pass looked nearly complete. Trucking still found defects, but different ones:

  • ⁠New missing components were rare.⁠ The screen mostly validated foundations in their natural jobs.
  • The largest remaining gap was ⁠an existing component hidden by the catalog⁠. Bulk assignment belongs in ActionDataTable through selectedIds and bulkActions, but the catalog sent the agent to plain DataTable. Dispatch fundamentally selects several unassigned jobs and assigns them to a vehicle, yet poor guidance made that the only hand-rolled workflow.
  • A narrow industry-specific gap remained in LimitMonitor, combining a value with soft and hard hour limits and an exceeded state. Meter plus ReferenceValue can compose it, but no dedicated component exists.

Once the component pass matures, the test shifts from producing new UI to correcting guidance that hides what is already shipped. Rankings pointing to StatGroup in , rating input pointing to inert Rating in , and assignment pointing away from ActionDataTable here all share the same active misdirection.

I corrected the catalog in PR#418 to send job lists with bulk assignment to ActionDataTable. I recorded LimitMonitor on issue #417 and added the second absolute-threshold case to issue #412. ⁠Nothing new was added under src this round.⁠

📊 Component scoreboard (⁠18 built⁠)

Gantt with intraday resolution / Leaderboard / SegmentedControl / LineChip, plus earlier components In progress: SectionList or TransactionList at two of three, LimitMonitor or an absolute Meter threshold at two of three, and ValidityTimer at one of three.

📋 Mode progress: the fifth mode begins

  • ✈️ Air ✅ / 🚆 Rail ✅ / 🚕 Taxi ✅ / 🚌 Bus ✅
  • 🚚 ⁠Trucking: first staff screen, dispatch⁠, opening the final mode

Next up: build log

  • Continue staff-facing trucking with load matching, freight billing, or fleet maintenance, then move to shipper pickup requests and recipient tracking.

The run

In the fifth mode, components returned to the work they were built for. Once the catalog matures, correcting guidance that hides existing components becomes the main result.

This series is co-created with AI (Claude and Codex). A human designs the experiments, makes the calls, and fact-checks before publishing; the AI does the hands-on work and drafting.

Components from @gunjo/ui

The components this screen's source imports directly.

The code the agent actually wrote

Click a file to expand its source.

業界別「AI指示書パック」、先行登録を受付中です。