We ship fast. Really fast.

A full market-assessment tool for agents: pull comparables, map them, drop in photos, and export a printable report — with autosave and versioning so nothing gets lost. Their spreadsheets can retire now.

A proper home for our agency products, from data enriched deeds for market assessments. Market overview dashboard built specifically for Agents.

We started mapping cadastral boundaries and municipal zoning across the Cape. The groundwork for knowing exactly what a property is — and what you can (and can't) build on it.

Nearly 1,500 listings from eXp Realty — the world's largest cloud brokerage — joined the platform. The roster keeps getting bigger.

A market overview for renters and investors: agency mix, rent spread, and how long listings actually sit before they let. The rental market, finally legible.

We built an MCP server and a ChatGPT app widget. Ask ChatGPT to find you a home and it actually can — results and all, right in the chat. In early access.

Bespoke property reports, sample galleries, and a request flow — built for the Reside Summit. Property intel, gift-wrapped.

Tell us what matters to you once, and your results lean that way. Settings that actually do something.

Self-service subscription cancellation. No emails, no hoops, no "are you sure" guilt trip. (Okay, one.)

Rotated to Supabase's new publishable and secret keys and retired the legacy service-role key across every edge function. Boring. Essential. Done.

Tell us where you work and when you need to arrive — we find homes within your real commute, traffic and all. Powered by HERE isolines because "arrive by 9am" was apparently too much to ask of anyone else.

Toggle between For Sale, Sold, and Rentals directly on the map. See what actually traded, not just what's listed. Filter, paginate, and explore in-view properties with full table views.

Rental listings landed, complete with projected yields. For the landlords doing the maths in the margins.

Multiple properties at the same address now stack neatly instead of fighting for attention. Click through them like a civilized person.

A dedicated AI property agent for teams, plus B2B pricing and an early-access waitlist. Same brain, business hours.

Repeat searches are now faster. We remember so you don't have to.

Filter suburb price charts by houses vs apartments vs townhouses, and by bedroom count. Finally, apples-to-apples comparisons.

Five agencies onboarded in a single month. Nearly 3,400 new listings added, including two of SA's most iconic property brands.

Complete rewrite of the search pipeline. Better structured query parsing, SSE streaming, and improved key counting for filters. The 13th time's the charm.

Properties without listed prices now handled gracefully across the entire stack — database, search API, and frontend. No more awkward blanks.

Sorting actually works now. Frontend, backend, the whole pipeline. Race conditions squashed.

If an agent claims a listing date that contradicts our records, we'll flag it. Trust, but verify.

Property pages and modals load significantly faster. We stopped asking the database the same question five different ways.

Hero, CTA, FAQ, the works. For when you need someone to help you afford the dream.

"Show me everything from [agency]" now works. They should probably thank us.

Endpoints for the AI Agent to search for properties, get location information, and get travel times.

We built a Model Context Protocol server. AI agents can now search SA property listings natively. The future called; we picked up.

"20 minutes from the office" now accounts for actual traffic. Departure time parameter added to travel time searches.

Switched primary login from magic links to OTP codes. Fewer "check your spam folder" moments.

Track properties you've interacted with. Your browsing history, but useful.

Built our own with Sanity CMS, custom styling, and sitemap integration. No more random blog links.

Deposit handling with percent/amount toggle. Math is hard; we did it for you.

Brand assets, logo downloads, and a proper about page. Two versions were made; one survived. Natural selection at work.

Reduced random logouts. Even we got annoyed with having to login again and again.

Four major agencies onboarded in a two-day blitz. Over 6,100 new listings hit the platform. The big leagues noticed.

Alerts actually work now! So you won't miss your dream property. And you can now unsubscribe from alerts!

When users search "modern home near UCT," we now use sequential semantic search for dramatically better results.

We know what you looked at, how long, and what you liked. Creepy? Maybe. Useful? Definitely.

Major framework upgrade with performance optimisations across the board.

Plans, tiers, and early adopter discounts. The "free forever" era has an asterisk now.

If we don't have your area yet, tell us. We're listening.

Queries without descriptive terms skip semantic search entirely. Faster, more accurate for "3 bed Camps Bay under 5M."

"Liked that search? Try these." Powered by AI, obviously.

Subscriptions, webhooks, payment processing. South African payments for South African property.

Unlocks sold property data, appreciation insights, and premium map features. Worth it.

Adaptive algorithm that adapts to what you're looking for. So that you get the best results possible without the search taking forever or never returning results. Possibily, the hardest thing we've ever built and our proudest achievement!

Investment insights report with sorting. Know which suburbs are actually growing.

Filter historical sales by date range. See what moved last quarter vs. last decade.

Seven agencies onboarded in October including ABC International, Link Property People, The Storey, and Lion Lead. The catalogue just got serious.

Ditched Google Maps for Mapbox. Better styling, better performance, better vibes.

Price, rooms, price per m² — all filterable directly on the map.

Bond costs, transfer costs, detailed breakdowns. Based on STBB data.

See what sold nearby, right on the property page map.

"A small slither of the image that has scrolled underneath was visible above the back button." The longest running, most annoying bug. Gone.

Five agencies in one month. Seeff and Chas Everitt alone brought over 8,700 listings. From scrappy startup to real coverage.

Full Mapbox implementation with bounds, markers, pagination, and mobile-responsive layouts. From zero to hero in a week.

Hover on the list, highlight on the map. Hover on the map, highlight in the list. It's the little things.

Date range pickers, ingestion monitoring, status tracking. Running a property platform requires knowing what's broken.

Locations parsed separately from descriptions using Google GenAI. "3 bed with pool in Camps Bay" now correctly identifies "Camps Bay" as a place, not a feature.

Results stream in real-time as they're found. No more staring at a spinner.

For-sale market analysis by suburb. Know what you're paying per square meter.

Hide properties you can't buy. Radical concept.

The first real listings hit the platform. Over 6,400 properties from three agencies. FindHomes went from prototype to actual property search.

See how a listing's price changed over time. Price drops highlighted.

Suburb-level market trends. Is Constantia getting more expensive? (Yes. Always yes.)

Government valuations on every property page. Compare what they think it's worth vs. what the market says.

Query quality checks before we waste compute. "asdfghjkl" is not a valid suburb.

Know your users. Ask them questions. Give them better results.

AI reads the images, understands the listing, and generates descriptions that actually help you decide.

Every property image analysed for features, quality, and content. The AI sees the pool you missed in photo 17.

Images served from the edge. Fast everywhere, not just in Cape Town.

Conversational property search. Ask follow-up questions. Refine as you go.

See what everyone else is searching for. FOMO as a feature.

App sidebar, new search experience, property modals, mobile-first everything. The version where it stopped looking like a hackathon project.

Type a suburb name, get suggestions. Basic, essential, finally done.

Client-side search with lazy-loaded image carousels and paginated results via search IDs.

Commit message: "Incredible!!" . Semantic search actually working. The moment everything changed. And we all thought the dangerous thought "How hard could it be?"

Desktop property photos in a Pinterest-style grid.

Reducer pattern to fix race conditions when applying filters. Goodbye, impossible filter states.

The AI understands bedrooms, price ranges, and pet-friendliness as discrete concepts.

Email notifications for new listings matching your search. We thought it was awesome but it wasn't... it was so inaccurate and we turned it off after a month. We needed a better search algorithm.

"START OF THE PROPERTY REVOLUTION" — actual commit message, August 17, 2024. Hybrid search combining keywords + semantic vectors. Auth, analytics, and the first property detail pages.

Keywords + semantic vectors. The portals use keywords. We use AI. There's a difference.

Every result shows how well it matches your search. Transparency is our thing.

Shadcn UI, Next.js, Supabase, and a bit of AI. We had no idea what we were getting ourselves into but just went for it.