All 60 of them, in one honest list. Filter by industry, who
it's for, or where it stands — the badge on every card tells you exactly
that, with no rounding up.
13 live·9 in pilot·11 in the works·8 weekend builds·15 paused4 retired— we show everything, including what we killed.
MeetNotes
Live
AI meeting notes in any language — speaker-separated transcripts, summaries and to-dos for the whole team.
Teams that want every meeting captured as clean, shareable notes without anyone typing
Shoppers comparing prices between UAE and India listings
buyiffy compared shopping prices across the UAE and India. It got as far as a working demo, and then hit the wall that kills most price-comparison products: keeping the catalogue accurate takes continuous, funded research. We stopped rather than let it quietly go stale. The traffic it did get turned out to be mostly bots, not shoppers.
Computer-vision safety monitoring for unattended baggage, fights, intrusion and falls.
Security teams that already run CCTV and want it to actually flag incidents
CameraUse watches existing camera feeds and flags what matters — unattended baggage, a fight breaking out, someone where they should not be, someone falling. It is paused, and not for lack of belief: it is a big project that needs proper hardware and a real sales motion, and we would rather say that than run a pilot we cannot support.
Sales forecasting for a clothing company — built for one client, never productised.
The clothing company it was built for
DataForecast predicted what a clothing company would sell, week by week and line by line, from its own sales history. It was built for one client, delivered, and never turned into something we sell. It is paused rather than retired because the forecasting work inside it is good and would transfer to any business carrying stock.
easyagent drove a phone by itself through Android's accessibility controls — tell it what you want and it taps through the apps. It is paused because that is exactly the capability Play's policy treats as dangerous. The version worth returning to is the one built for elderly users, where the same machinery is the point rather than the risk.
Describe a post, get a finished Instagram image and caption back.
Sellers and small brands who need a post today and have no designer
EasyInstaPost turned a sentence into a finished Instagram post — image, caption and hashtags, in every format Instagram wants. It never shipped on its own: VideoGuru became the creative engine and this folded into it. Paused rather than retired, because the 'one line in, finished post out' idea is still the right one.
Cricket prediction and value analysis — an independent fair win-probability, and whether the market price disagrees with it.
Cricket followers who want a calibrated read on a match rather than a tip
MasterOdds works out a match's fair win probability from 25 years of cricket data, then compares that with the price the market is quoting and says whether the market is wrong. It is an analysis engine, not a betting operator — it never takes a bet. It is paused: the model is built and backtested, and what it needs is a decision about whether we want to be in this space at all.
An AI lab-report analyzer that tracks health trends across a family.
Families who want their lab reports read in plain language and compared over time
MediTrack read lab reports and tracked a family's health trends over time — the numbers that only mean something when you can see them move across years. It is paused. Holding other people's medical reports is a responsibility we are not taking on casually alongside everything else.
Forward product photos on WhatsApp and get back a tappable shop link.
Social sellers who sell over WhatsApp and want a real checkout link
OnlyWAO turned a product photo forwarded on WhatsApp into a tappable shop link, so a social seller could sell without a storefront. It was started in July 2026 and put down: the same WhatsApp-first selling idea is better served inside the products that already have sellers on them.
phonebrain ran the model on the phone itself, with nothing leaving the device. We tried the local models that would fit and they were not good enough to stand behind. Paused until they are — and they are improving quickly.
Refplat was a two-sided marketplace where a startup or brand posts what it can offer, and micro-influencers who like it post about it and collect the reward. It is paused on the cold-start problem — both sides have to arrive at once — and there are others in that space now, so it needs a fresh look before it gets another go.
An admin-managed WhatsApp chatbot SaaS for event and booking partners.
Event and booking businesses that want a managed WhatsApp bot without building one
Samchatter gave event and booking partners a WhatsApp chatbot their own admin could run — the partner edits what it says, their customers just message it. It is paused: the assistant work that mattered went into Sam PA and the voice products, and nobody was pushing this one.
Write anything down and it files itself, keeping the date it was written.
Anyone who captures notes faster than they organise them
SmartNotes let you drop any scrap of text and filed it for you — into a bucket that already existed or a new one it made, asking when it was not sure, and always keeping the day you wrote it. It folded into Sam PA, where the same capture happens in a WhatsApp message. Paused as a standalone app.
An Android assistant that watched your screen and talked you through it — guidance only.
Anyone stuck in an app who would rather be told what to do than search for it
Ultiassist streamed a phone's screen and microphone to an assistant that talked the user through whatever they were stuck on — and deliberately never tapped anything itself. That restraint is what made it safe, and the same capability now lives in SamSupport and Sam PA. Paused as a standalone assistant.
VisionMnemonics was a weekend experiment in AI-generated visual mnemonics — turning a thing you need to remember into an image absurd enough to stick. It never grew past the experiment, and it is paused rather than retired because the idea keeps coming back.
AI daily digests of WhatsApp groups and DMs for busy operators.
Operators drowning in WhatsApp group and DM volume who needed a daily summary
whatsum worked and had a paying client, but the infrastructure cost of summarizing WhatsApp traffic at scale exceeded what it earned. It was retired for that reason, not because the product failed — a clean case study in unit economics.
MovAuto was a near-duplicate of EasyInstaPost, built before we admitted the two were the same product. Both folded into VideoGuru. This one is retired for good rather than paused: there is nothing in it that EasyInstaPost's own entry does not already cover.
An internal ticketing system for service issues — roles, lifecycle, and an analytics view.
Internal team only
NeoService was an internal service-desk build: tickets with a real lifecycle, agent and manager roles, and a dashboard over the whole thing. It never had an outside user, and it is retired rather than paused — every business it would serve is better off with software that already exists.
AI Google Sheets analytics and PDF-to-Excel conversion.
Anyone who needs to turn a messy PDF into a working spreadsheet or ask their sheet a question
SheetBoy answered questions over a Google Sheet and turned PDFs into working spreadsheets. We retired it: general AI chat now does most of this, and it had one real user. People still ask for it back for the thing chat is bad at — 150,000-row sheets and fast search over them — so the idea may yet return in something else.
A sticky-note work board — five columns, hard caps on what you can call urgent.
Personal use only
WorMa was a personal kanban board with one opinionated rule: the two 'do this first' columns were capped at five cards, so declaring everything urgent was physically impossible. Built for one person, never intended for anyone else, and retired.