Brainstorm word cloud
Terms from this hackathon's own brief, judging criteria, discussion board and submissions, plus our guide. Bigger = mentioned more often; wording that most Devpost hackathons share is left out. Click a word to see past winning projects that mention it.
Key facts
- Two tracks, one submission
- General Track: any working software or hardware project, any topic. Theme Track, Rethinking Waste: a project whose main purpose is reducing, sorting or recycling waste. Each team submits once, to one track, and you may switch track before the deadline. The tracks are judged on separate rubrics for separate prizes ↗
- The theme must be central
- Judges check that waste is the point of the project, not a label. The organizers' own example: a to-do app with a "recycling" label counts as a General project, not a Theme one ↗
- What to submit
- A public code repository (GitHub, GitLab or similar), a demo video of 2 to 3 minutes at most on YouTube or Vimeo showing the project working, and a written description of what it does, how you built it and what you would do next. Theme entries add one paragraph naming the waste problem and how you would measure whether it worked. No repo or no video makes the entry ineligible, and late entries are not judged ↗
- Who can enter
- Students aged 13 or older, solo or in teams of 1 to 4. Companies and professional organizations are excluded. All countries are open apart from Devpost's standard exceptions ↗
- How much of the prize is cash
- The overview lists 1st place in each track at $2,100: $100 cash, $1,000 in Adaption credits and $1,000 in Deployxa credits. 2nd place in each track is $1,550: $50 cash, $1,000 Adaption and $500 Deployxa. A participation prize is $500 in Adaption credits. Together that is the $7,800 headline, of which $300 is cash ↗
- The rules page lists different prizes
- The rules page still shows every prize as TBD and names extra categories that are not on the overview: Most Social Impact and Best Pitch for General, Best ML Usage and Best Pitch for Theme. Ask on the organizers' Discord which list is final ↗
- Open questions on the forum
- On Oct 17 an entrant asked whether a prototype started before the Oct 20 start date may be submitted if disclosed, and whether AI coding assistants are allowed. As of Oct 27 the thread had no reply, and neither the overview nor the rules say anything on either point ↗
- What wins in waste projects (our count)
- On Oct 27, 2026 we read the first 300 Devpost search results each for recycling, food waste, e-waste and compost. Of 1,113 unique projects, 496 mention waste, recycling, compost, trash or sorting in the name, tagline or tags, and 33 of those carry a Devpost winner badge (6.7%). The badge covers any prize at any hackathon, so treat this as a pattern, not a rule ↗
- Hardware and team size in that sample
- 62 projects list hardware tags such as raspberry-pi, arduino, esp32 or servo, and 10 of them won (16%), against 23 of the other 434 (5.3%). By people listed on the project: solo 7 of 244 won (2.9%), two people 4 of 86 (4.7%), three 9 of 77 (11.7%), four or more 13 of 82 (15.9%) ↗
- Which ideas are crowded
- Sorting and "which bin" apps are 197 of the 496 projects (40%), with 17 winners (8.6%). Food waste apps are 139 with 7 winners (5.0%), composting 32 with 1, e-waste 17 with 1. Sorting wins often, but you are up against the most look-alikes, so the 25% for Innovation in General or the 35% for Impact in Theme depends on what makes yours different ↗
- What Skill Rank shows
- None of the 1,074 entries tracked on PocketPlay Skill Rank on Oct 27, 2026 mention recycling, waste, YOLO, OpenCV, Raspberry Pi or Arduino, so there is no ready-made agent skill for this theme ↗
Idea angles
Our scoping ideas for beginners, built from the official brief and our Devpost count. They are not official briefs.
Plain photo-to-bin apps are 40% of the field. Two winners in our sample went further: one judges an item by its condition (greasy, broken, still full) and one handles the fact that recycling rules differ by place. Pick one real set of local rules, such as your school or town, and ask the follow-up question a photo cannot answer.
Three food-waste winners in our sample predicted how much food to prepare from past consumption, and two of them list pandas and scikit-learn. A forecast for one cafeteria, with a number for kilograms not cooked, speaks straight to the 35% Impact and 20% Feasibility rows.
Hardware entries won about three times as often in our sample. A Raspberry Pi, a camera and one servo that flips a lid is enough for a demo, and the physical sorting reads clearly in a 2 to 3 minute video.
The Theme paragraph asks how you would measure success. Build the measurement into the project: log what went into which bin, show the wrong-bin rate per day, and put that chart in the video.
General weights technical execution at 35% and polish at 25%. Choose a project small enough to finish and harden, then spend the last days fixing broken flows instead of adding features.
Past winners in related hackathons
Winner-badged projects from ended Devpost hackathons with the same theme that mention terms from this brief, closest first. Read a few before you pick an idea.
- Leaf It Off — Reducing energy waste in college dorms with one cute plant at a time. (Girls Hoo Hack 2026)
- Playground — Snap a photo, transform physical objects into playable 2D games, and join a global platform! (SummerHacks)
- LandMarked — Claim real places near you with a single photo, the place keeps every photo forever, and together they answer a strange question live: what colour is this neighbourhood right now? (SummerHacks)
Judging criteria
Each track is scored out of 100 on its own rubric. The first four rows are the General Track, the last four the Theme Track. Weights are from the overview and the rules page, which agree. source ↗
Code quality, working functionality and how hard the build was. This is the largest share in General, so a working, non-trivial build beats a pretty mockup.
How novel the idea or approach is.
How finished it is. Rough edges, broken features and unfinished flows lose points, so cut scope before the deadline rather than leave half-built screens.
How clearly the video and write-up explain what was built.
How much the project reduces, sorts or improves recycling of waste. The required paragraph on how you would measure success is your best evidence here.
Code quality, working functionality and difficulty, as in General.
Could it be deployed in the real world? The organizers say a working solution beats a clever one that cannot ship.
How clearly the video and write-up explain what was built. Worth more here than in General.
Starter stack
- Teachable Machine Google's browser tool for training an image model without code, and the tool one winner in our sample used. Good for a first prototype of a sorter in an afternoon ↗
- Ultralytics YOLO quickstart Install and run an object detection model in Python, for a camera that finds several items in one frame. YOLO was a tag on two winners in our sample ↗
- TrashNet dataset 2,527 photos in six classes: glass 501, paper 594, cardboard 403, plastic 482, metal 410 and trash 137. Every photo was taken on a white posterboard, so test on photos of real bins before you trust the accuracy ↗
- TACO dataset Litter photographed in the wild, in woods, on roads and beaches, with COCO-format annotations for detection and segmentation. The paper describes 1,500 images and 4,784 annotations ↗
- Arbiter Discord The organizers post all updates there, and it is the place to ask about prizes and the open forum questions ↗
Timeline
- Hackathon opens (12:00 pm EDT)
- Submission deadline (9:00 pm PST, midnight EST); late entries are not judged
How a beginner can ship
- Register and pick a track
Join on Devpost and choose General or Rethinking Waste. You can switch before the deadline, but you can only submit to one.
- Ask your open questions early
Join the Discord and ask which prize list is final, whether earlier work counts and whether AI coding tools are fine, then keep the answers.
- Make the repo public from day one
A missing repo makes the entry ineligible. Add a README with run steps as you go.
- Get one loop working
For a sorter: one photo in, one bin out. For a forecast: one week of data in, one number out. Build features only after that works.
- Write the measurement paragraph (Theme)
Name the specific waste problem and how you would measure success, such as wrong-bin rate or kilograms of food not cooked.
- Record a 2 to 3 minute video
Show the project working, not slides. Upload to YouTube or Vimeo and link it.
- Submit before Dec 20, 9:00 pm PST
Repo, video and description. Late submissions are not judged, so submit a few days early and update the draft.