The problem Giam Siap solves

Problem Statement

Three related pains, one system:

• Owners waste real time and money just watching prices move. Ingredient prices shift day to day across vendors, but checking them is manual and unrewarding work — so owners either miss the dip and overpay, or react too slowly and stock out mid-service.
• SME restaurants and cafes don't have a procurement team to fall back on. A larger F&B chain has staff whose job is to know what a fair market price looks like and how much stock the business actually needs. A single-outlet SME owner has none of that — they're guessing at both numbers themselves, on top of running the floor, which means the business either overpays for what it buys or over/under-orders against demand it never had time to properly size.
• Every order today means a negotiation, and every negotiation means friction on both sides. Owners spend time haggling price and credit terms with suppliers instead of running their business; suppliers spend that same time chasing Net-30/Net-60 invoices instead of getting paid, which strains their cash flow and ties up working capital that should already be theirs.
• Neither side can safely automate any of this today. Web2 tooling can't be trusted with real spending authority without either a human re-approving every transaction (which defeats the point of automating it) or a high-fee third-party processor sitting in the middle (which eats the margin automation was supposed to protect).

Giam Siap's answer, in one sentence: it gives an SME owner the procurement team they can't afford to hire, and replaces the negotiation both sides can't afford to keep wasting time at.
The owner doesn't need to arrive with a price and quantity already worked out — they can just state a need ("I want to procure coffee beans"), and the agent asks a few follow-ups about their actual business (location, operating days, typical volume, stock on hand), researches the real current market price itself, and proposes both numbers with its reasoning, doing the market research and demand sizing a bigger business would pay a team to do. And because the owner is only ever agreeing to an objective, pre-committed condition — not haggling a deal with a person on the other end — there's nothing left to negotiate: the smart contract, not the agent and not any human, locks the funds the moment the owner confirms and releases them the instant the vendor's own signed price meets that condition, so the supplier is paid instantly and trustlessly, with no invoice to chase and no counterparty risk on either side.

Challenges we ran into

Challenges We Ran Into

No real vendor to test settlement against. Early on, the biggest blocker to proving the escrow actually worked end-to-end was that we didn't have a real, registered vendor issuing signed price quotes — and the contract's whole trust model depends on verifying exactly that signature before it will release funds. Without one, we couldn't exercise the release path at all, only the lock path. We solved this by standing up a mock vendor endpoint that behaves like a real registered vendor: it signs price quotes the same way a genuine vendor integration would, so checkVendorPrice and the watcher's settlement logic could be exercised against something structurally identical to production, not stubbed out. That let us validate the full escrow → price-match → payout flow was technically sound before a single real vendor was ever onboarded.

Getting owners in is easy — getting vendors in is the hard part. zkLogin means an owner can join in one Google sign-in with no wallet, no gas, no crypto knowledge. Vendors don't get that same free ride: settling a real trade means a vendor has to actually be reachable on-chain, which is a much bigger ask than "log in with Google" for a supplier who's never touched a blockchain and has no reason to care that they're on one. This is really a two-sided cold-start problem — the buyer side of the marketplace is nearly frictionless, but the supply side isn't, and a marketplace with only buyers doesn't work.

Longer-term, our plan is to make joining worth a vendor's while directly: a commission/toll-fee reward for vendors who come on-chain, with the reward gated behind a minimum retention period (e.g. staying active on-chain for 12 months) so we're paying for vendors who actually stick around and keep settling trades, not one-off sign-ups chasing a payout.