A reverse-chronological record of meaningful updates, feature releases, and fixes.
Published the full dataset as open data on GitHub: 21 routes, 218 stations, and 217 route segments — every fact traced to its source (31 references, roughly 1,240 citations), plus historic bridges and tunnels. It’s a point-in-time snapshot generated straight from our database and released under the Open Database License, with OpenStreetMap attribution for the real track geometry. Anyone can browse it, cite it, check our work, or build on it — and it’s reproducible by design: a single script regenerates the whole thing from our database.
Gave the site a real landing page with a short, swipeable walkthrough of the project — the vision, the history it draws on, the honesty-first map, the sourcing, and how to get involved — fronted by a playful ‘someday’ boarding pass (Cairo→Beirut, and a few other dream routes). Rebuilt the public map on our own database, retiring the old embed, so it now shows real track, best-guess corridors, and every feature’s sources. Along the way: a fuller footer that maps the whole site, a historical timeline now running from the 1850s to today, and a sweep of typography and spacing fixes so everything reads cleanly.
Big mapping upgrade, now live. Made the new research catalogue the app’s source of truth — 21 routes and 218 stations, with every fact cited back to a source — then recovered real historic track geometry from OpenStreetMap for 85 route segments: the Barada gorge, the Lebanese coast, the Jordan Hejaz, the Egyptian Nile valley and more. The Lab map now draws the network to the best of our knowledge — solid where the track is real, dashed or dotted where it’s still a best guess, with bridges and tunnels marked — and every feature shows its sources.
Rebuilt the rail dataset from the ground up: 21 routes and 218 stations spanning 1854–1992 — and this time every attribute traces to a cited source, with two honesty axes tracked throughout (how precisely each station’s location is known, and how well each route’s path is known). The review caught real errors in our earlier data — a couple of swapped gauges, plus some out-of-order and duplicated Hejaz stations — and added a layer of historic bridges and tunnels. Done with Claude’s research tool.
Added two public housekeeping pages: /about/log (this page) as a reverse-chronological record of what’s shipping, and /about/specs as a living inventory of the open-source libraries, AI tools, and data sources that power Levantrain.
Rebuilt the site off the legacy Vite single-page app onto the current Lovable stack (TanStack Start + Tailwind v4), on a Supabase backend we run ourselves. Two reasons drove it: the old SPA couldn’t do proper per-page SEO or social-share cards — server-side rendering fixes that, so sharing a blog post now shows the right title, description, and image; and we wanted our data on infrastructure we fully control rather than a managed backend, so we can grant access to third-party tools and agents as the project grows. Along the way, all 33 pages became real file-based routes, the domain was cut over to the new deployment (blog, avatar, and resource images migrated first), the code was consolidated under the Levantrain open-source org next to the dataset, and the old codebase was archived. Unglamorous plumbing, but it keeps us independent and easy to build on in the open.
Ran a series of deep, fact-checked research passes on Middle East railways — the history, what runs today, and what’s being planned — and turned the findings into a first machine-readable geodata kit: a source-cited registry of lines and stations, plus a pipeline to pull real track geometry from OpenStreetMap. Groundwork for a map that’s honest about what we do and don’t yet know. Done with Claude’s research tool.
Shipped a full blog — posts, authors, and an RSS feed — with dynamic social-share cards that render a preview image per post. Also wired up continuous integration with unit and end-to-end tests, so changes get checked automatically before they ship.
Built the tools that maintain the data: an authenticated admin for stations, routes, and resources — including a map where you can draw a route and drag its stations into place — and a public “Lab” to browse the same data openly. The first interactive map (v0.1) went live on OpenStreetMap.
The big build week. Stood up a proper database, authentication, and file storage, and modeled the network relationally — stations as the single source of truth, routes as ordered sequences of them. Compiled a first structured dataset of ~85 historic Levant stations and their lines, hand-gathered from archival maps, timetables, and references, and published it as an open-source repository so anyone can use or improve it. Set up our GitHub org along the way.
Opened up the community and social channels — a Google Group for discussion, plus LinkedIn and Bluesky for updates — so the project had somewhere for people to follow along and weigh in.
Put up the first real website: a landing page explaining the idea, a page on the region’s railway history, a roadmap, and a note from the founder on why a peaceful, connected Middle East is worth imagining out loud. The public front door for the project.
Started digging into the basics — did a rail network like this ever exist, what’s still standing, and had anyone tried to map it? (Short answer: much of it did, some still stands, and no one had put it together openly.)
Levantrain began as a thought experiment back in June 2024 — starting, fittingly, as a long conversation kicking the idea around: if the Middle East were ever at peace, one of the things it would need to thrive is a modern rail network connecting its people — the kind of link that a century ago almost ran from London to Cairo. Rather than wait, why not start imagining and mapping it now?