The server locked the scores. The judge’s screen still said “Update.”
Glassbox had a misleading screen at exactly the moment a judging platform needs to be clearest. An organizer published the results. The server correctly refused further score changes. But the judge’s page still looked editable. A judge could spend time revising a comment, press Update, and discover that the action had already become impossible. This turned up during the build’s browser…
During a testing phase for the DOGFOOD 2026 hackathon, a bug was discovered in the Glassbox judging platform. The problem occurred when a judge's screen showed an "Update" button, despite the system already being locked to prevent further score changes. This issue arose while testing the browser's response to the transition from a modifiable state to a final, published state. The fix for this bug was relatively simple, involving changes to only one commit in the codebase.
The Glassbox platform is an open-source, self-hosted solution designed for submission and judging purposes. It's built using Node.js 24, TypeScript, SQLite, and server-rendered HTML, without any runtime npm dependencies. The most important feature of Glassbox is its "Publish" button, which finalizes the judging process. Before publication, judges can edit scores, and organizers can view raw and calibrated rankings. After publication, the focus shifts to presenting a stable result to the public.
The bug lay in the browser's failure to register the transition from an editable state to a locked, final state. To address this, the developers altered the locked state of the score form based on the event phase. Upon successful publication, the page displayed a locked notice, disabled the score fields, and removed the save and recusal actions. However, the backend still had to reject any forbidden writes, ensuring the integrity of the system.
To verify the fix, a regression test was conducted. This test involved opening an assigned score page as a judge and confirming that the save function was available before publication. After publishing the event, the judge's page was reloaded, and the test checked for the presence of a locked notice, disabled fieldset, and the absence of save and recusal controls. Both the rendered HTML and the backend's rejection of forbidden writes were inspected during this test.
The implementation of Glassbox includes a ranking system that combines a judge's individual scores with an event-wide average, using pseudo-reviews by default. This approach ensures that sparse judging histories receive more weight in the final ranking. The system also preserves the raw ranking, allowing organizers to compare adjusted and raw rankings, inspect judge offsets and spreads, and export review-level values.
While Glassbox provides a robust solution for judging and ranking, it is essential to remember that it remains a single-node SQLite application without community voting, email delivery, or built-in login rate limiting. The hash chain used in Glassbox detects inconsistent edits, but there is no external trust anchor to guarantee the integrity of the system.
The committed acceptance report for Glassbox includes seven passing checks covering various aspects of the platform, including public gallery behavior, deadline rejection, score access isolation, and CSV export. However, these checks do not establish production readiness or real-world fairness.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
