All projects

Food trends usually blow up overseas first. Lead Time watches weekly search interest for 22 foods across seven markets, works out how long each one has historically taken to reach Singapore, and tells a small food business which trends are heading its way. For each food it gives a verdict (Go, Go but late, Wait and watch, or Skip), checks whether the kitchen can actually make it, and then helps with the launch: a suggested price and margin, when to bulk order, a launch offer with its break-even, and a dated posting plan.

Runs on a free server, so the first load can take up to a minute.

Say or type what you feel like eating, in English, 中文, Bahasa Melayu or தமிழ், and Voucher Kaki finds CDC-voucher eateries near you. Results are ranked by relevance and walking distance, with a deliberate nudge toward independent shops over chains. When it has to widen the search to fill the list, it tells you what it gave up instead of quietly showing worse matches.

People are starting to shop by asking an AI assistant instead of typing into a search box. A question like “lightweight shoes for a half marathon in humid weather, under S$200” needs facts most product pages never state, so those products quietly never get recommended. AgentShopper takes a catalogue file (CSV, Excel, Shopify JSON or even a scanned PDF), scores how ready it is for AI shoppers, lists exactly which facts are missing, and writes content for each product: who it suits, who it doesn’t, and how it compares with the real products beside it.

The two built-in demo catalogues, running shoes and facial skincare, work without uploading anything. The app sleeps when idle; press its wake-up button and give it a minute.

Investment scams were the biggest single source of scam losses in Singapore in the first half of 2026. They rarely start with an obvious lie. They start with a friendly “personal manager”, a group chat full of people posting profits, and a small first deposit. ChainCheck takes what the offer showed you, as pasted text or screenshots, and checks it: is the firm on the MAS register, is the website a week old, do the promised returns make sense over a year, is there a fee to get your own money out. Then it shows which stage of the playbook you’re in and what usually comes next.

The MAS register on the demo uses invented firms and is labelled as demo data on screen, so a real firm is never called unlicensed. Free server, so the first load can take up to a minute.

Most food waste at home isn’t one big mistake. It’s the yoghurt found three days late, and the fourth tin of chickpeas bought because you forgot the other three. Fridge Rescue treats those as two problems. Perishables close to their date go into one recipe that uses all of them. Unopened, long-life surplus goes into a donation box, with the nearest drop-off point. Sprout, a pixel pet on the home screen, thrives when you rescue food and wilts when it goes in the bin.

The Android APK and a 1:34 demo video are linked from the repo. In the browser, the purchase screen is a preview.

A serum costs $30 to $120 and takes 4 to 12 weeks to show anything. By the time the bottle is nearly empty, you have no baseline, no way to tell a real change from different lighting, and no way to tell “it isn’t working” from “I kept forgetting to use it”. Glowline is built for that moment: deciding whether to buy the same bottle again.

Each concern has its own response window, so a retinol isn’t judged on day three. Free server, so the first load can take up to a minute.

Every repo has a graveyard: functions that were deleted and never came back. git seance walks the history, finds them, and lets one speak, with who wrote it, when, and the commit that killed it.