I lost 3 hackathons in a row. So I built a vampire judge to tell me why.

Description and use

The author built Judge Fangs, an AI tool that evaluates hackathon submissions against winning patterns, providing scores, reasons to lose, and actionable fixes.

Problem

Repeatedly losing hackathons without clear feedback on why the submissions were not winning.

Outcome

Judge Fangs scored 9 of 10 real winners in its top 10, with an average winner score of 96 versus 64 for non-winners, demonstrating alignment with real jury judgments.

Workflow

• 1. Provide the hackathon challenge URL and the submission (DEV post, GitHub repo, email, or idea).
• 2. Judge Fangs fetches the challenge page, winners announcement via DEV API, and the repo via GitHub API; if winners post missing, uses SerpApi to locate it.
• 3. Gemma extracts judging criteria and patterns common to winners.
• 4. A Python script checks hard facts: presence of claimed technologies, title elements, demo links, video length, etc.
• 5. One call to Gemma judges the submission, returning a JSON with scores per criterion, a verdict, evidence quotes, top 3 reasons to lose, 3 fixes, and improved title and opening.
• 6. The tool presents the verdict, reads it aloud, and displays actionable feedback.

Evidence

Judge Fangs scored all 45 without being told which ones won. The "what winners have in common" step was built from other winning posts, so he couldn't peek at the answers.

Source

The "what winners have in common" step was built from other winning posts, so he couldn't peek at the answers.

Source

9 of the 10 real winners landed in his top 10.

Source

Average score of winners: 96. Non-winners: 64.

Source