Baek MinHyeok@MIN_
1
Events attended
1
Submissions made
Korea, Republic of
1 year of experience
About me
AI-native indie developer from Korea I build products through conversations with AI — shipping fast, iterating faster. This approach led me to launch multiple apps including a subscription manager (SubCare), a price-to-unit converter (그돈씨) Currently obsessed with agents, automation, and finding the fastest path from idea to working product. Always hacking. Always shipping.
🤓 Latest Submissions
HypeMeter: Score the Hype, Show the Evidence
Every week, companies announce bold claims — hiring sprees, partnerships, explosive user growth. For investors and journalists, verifying a single claim against reality can take hours or days of manual digging. There is no fast way to separate substance from hype. HypeMeter closes that gap. Paste any corporate claim — an IR statement, an earnings line, a press-release quote — and an AI agent gathers live, publicly observable web evidence to compute a Hype Score (0–100): the measurable distance between what a company says and what the public web actually shows. Under the hood, a Gemini 2.5 Flash agent (via Function Calling) autonomously queries multiple live sources through Bright Data's SERP API — LinkedIn Jobs, Glassdoor, Reddit, and news — then cross-references the numbers. Tesla's "hiring 3,000 AI engineers" claim scores 60 / Partially Supported: against 174+ observable open AI engineering roles, a real and transparent gap. Every score links back to its sources. It works on history too — run it on WeWork (2019) or Nikola (2020) and it flags the same gaps the market discovered far too late. For hedge funds, retail investors, and financial journalists, this is due-diligence triage at the speed of a headline — an analyst's week of digging in 30 seconds. Every result shows its sources and scoring method, never a black-box number. HypeMeter is built to be legally careful: it measures an evidence gap between public claims and observable evidence — it never accuses any company of wrongdoing, and is not financial advice. Built with Bright Data (SERP API — the evidence engine), Gemini 2.5 Flash, Next.js 15, Vercel, and Neon Postgres. Built solo in five days for the Web Data UNLOCKED Hackathon.
31 May 2026
