-
Synthetic demo: 8 matched hours; 0.960 kWh before, 0.360 after, 0.600 difference. Generated data, not measured savings.
-
Synthetic increase: 100W before, 110W after across 8h gives -0.080 kWh difference. No measured savings or causal claim.
-
Synthetic gap test: 8h before, 4h after. Shared 4h use 0.480 kWh each: zero difference. Unmatched hours excluded.
IdleProof Find the electricity that stays on — and keep the evidence honest.
Inspiration
An overnight power reading does not tell us which equipment is unnecessary, and a lower bill does not prove a shutdown caused the change. IdleProof turns a small, participant-run before/after experiment into an inspectable comparison, with visible gaps and no inflated annual or carbon claim.
What it does
Import two nights of interval-average electricity measurements, choose the night window and compare only the clock-time coverage both nights actually contain. The chart, totals, missing-hour disclosure and downloadable evidence JSON update together. The page runs locally and never sends the imported readings to an API. The initial screen is a clearly labelled synthetic demonstration, not measured environmental savings.
How we built it
Original HTML, CSS and JavaScript ES modules. A pure calculation module validates CSV records, clips intervals to a fixed-offset local night, intersects coverage, and integrates average watts over matched hours. Browser-native SVG renders real calculated data. An exported JSON includes settings, matched intervals, limitations, the source CSV and its SHA-256. No remote model, paid service, framework or external dataset is required.
Challenges
Missing intervals can make a naive before/after total look better than it is. IdleProof removes unmatched time from both sides rather than treating missing data as zero. Overlapping readings, ambiguous offsets, DST transitions, invalid values, long intervals and multiple nights fail closed. A mobile review found tiny chart labels; a separate compact chart layout fixed their legibility.
Validation
18 automated numeric/data-quality tests pass, including asymmetric missing coverage, partial-window clipping, equivalent subdivisions, increased after consumption and invalid timestamp/overlap cases. Native browser checks verified real file import, recalculation, invalid-data blocking and an actual evidence download; its source hash and numerical results were checked. A separate central reviewer inspected the exact six v003 source files, claims, secret scan and 20 tests (including 200 irregular-interval cases). This is local source review, not organizer eligibility approval or a field study.
Environmental benefit and limitations
Intended users are students or small shared spaces with permission to review interval-average meter data. The prototype supports checking whether a carefully chosen shutdown experiment coincides with lower overnight consumption. No real electricity savings, emissions reduction, causal effect or user adoption have been measured. Weather, occupancy, equipment changes and meter accuracy can confound the comparison. Never disconnect refrigeration, medical, safety or essential equipment based on this tool. Public source and runnable instructions: https://github.com/estona815/idleproof-green-shift-2026 . A public step-by-step walkthrough is included in its README. A hosted live demo is not provided. Individual event registration is complete; final project submission was completed on October 20, 2026.
AI and asset disclosure
OpenAI Codex assisted with concept, original code, synthetic examples, documentation and producer review under participant kwonj0815's instructions. A built-in image-generation tool produced a separate design reference, not a functional screenshot masquerading as UI. No prior project code, stock assets, external fonts, music, third-party dataset or paid API is incorporated. Organizer approval of AI assistance is not claimed.
Log in or sign up for Devpost to join the conversation.