Hangry is the deliberate pivot from Onz, which I closed in June 2026. Onz built an AI recommendation engine before it had the users whose data the engine needed. Hangry takes the same problem — decision fatigue in food — and inverts the order of operations: collect the supply by hand first, write the product second. I run it out of Kaizer Studios as founder, doing the product design and the restaurant conversations myself.
An independent operator with a genuinely good signature dish has no way to be found for it. A great matcha or a standout satay gets buried under a generic map pin or a forgettable Instagram post. Every existing route to visibility costs the operator something they did not choose: a discount set by a deals platform, a commission on delivery, or marketing work they have no time to do. The gap is not that these restaurants are undiscovered — it is that they are discovered as a location rather than as a dish.
Eighty venues tracked, twenty-one emailed, six onboarded. For cold outreach that funnel is unremarkable. The composition of the six is the finding: the partners who say yes are cafés, not kopitiams — four of the six are cafés, one is a local restaurant. Not one kopitiam or hawker has onboarded, and those are exactly the venues a nasi-lemak-first launch depends on. A café already thinks in brand, photography and Instagram, so a curated dish listing reads as free marketing it would otherwise pay for. A kopitiam has none of that frame, does not discount, and does not feel the absence of another channel. The launch thesis and the supply that actually converts are pulling in opposite directions.
So on 28 June I dropped the onboarding gate and decided to launch scraped-first — I control the supply outright through manual collection, every listing carries a visible provenance flag and a last-verified date, and onboarding runs alongside launch rather than in front of it. That reverses the sales conversation: instead of asking an owner to imagine a listing, I show them one that already exists, already ranks, and is one “claim this listing” away from being theirs to edit.
Scraped-first buys supply at the cost of consent. A listing an owner never asked for is a weaker starting position than one they signed up for, and the listing agreement that would formalise it is still an open item rather than a shipped one. I took the trade because a directory with six venues does not work and a directory with eighty does — but it means the first conversation with an operator is now about something already published, not something proposed, and I have to be able to defend that in the room.
The promotion loop is the whole ROI argument I make to a restaurant — list them as their proxy website, we distribute it, you see footfall. Until that number moves, the pitch is a hypothesis with six data points, not a proven mechanic. Pricing is untested for the same reason: nobody has been asked to pay yet, so what an owner will actually part with is still unknown. And the September gate is twenty onboarded by the fifteenth. At six, with twelve days left, the honest thing is to present the gap rather than the target.