A new platform called Kedge lets developers run applications across multiple regions of the world, with built-in data storage and a billing model that charges by the second for what you actually use. It launched on Hacker News in a post by its creator, who previously spent four years helping build Fly.io, a well-known cloud hosting service. kedge.dev

Most cloud platforms that offer "serverless" computing — where you run code without managing servers — treat each request as independent and store any needed data in a separate database. Kedge takes a different approach: it keeps data built into the app itself, copied across regions, accessible through SQLite, a popular and familiar database tool that runs as a simple file. The creator cited an essay called "The Serverless Server" on Fly.io's engineering blog as the inspiration. kedge.dev

Kedge can start up a new virtual machine — a small, self-contained computer environment — in 3 milliseconds. It does this by keeping a set of pre-started machines ready and cloning them when needed, sharing memory between copies until one actually needs to make a change. The coordination layer that manages all of this runs on SQLite, which is an unusual choice; most platforms use specialized systems designed to keep multiple servers in exact agreement.

Kedge gives apps two built-in tools for storing data: a shared database accessible from any instance, and a shared file system. Both are powered by syzy, an open-source tool that lets multiple copies of data stay in sync even when they are edited at the same time, using a technique called CRDTs (Conflict-free Replicated Data Types — data structures designed so that simultaneous changes can be merged automatically without anyone having to resolve conflicts manually). It is available at github.com/wjordan/syzy.

Beyond running code, Kedge supports building web applications directly. A working clone of Hacker News, complete with story submission, voting, comments, and login, was built in roughly 60 lines of Markdown plus CSS — a strikingly small amount of code for a fully functional app. It is documented at kedge.dev/docs/html-apps#kedger-news. The platform also supports static sites, serverless functions, full-stack apps, and databases placed close to users. kedge.dev

The billing model is granular. Users are charged per second on actual resources consumed: CPU charges apply only for milliseconds of compute time actually used, memory charges track active data, and storage charges bill only for bytes used. Listed rates are $15 per vCPU-month, $5 per GB-month for memory, $0.05 per GB-month for storage, and $0.01 per GB for egress (data sent out from the platform). Idle apps scale to zero and pay only for storage. The platform uses a credits-based billing system with a $5/month free tier. Users can log in via GitHub OAuth or by running ssh kedge.dev from a terminal. kedge.dev

At the time of the initial launch, Kedge's preview ran in 11 regions with no billing system yet implemented; the pricing page was described as an estimate. The current kedge.dev site, dated July 29, 2026, now lists concrete rates and the credits-based free tier, suggesting the billing infrastructure has matured since the initial launch. Hacker News kedge.dev

The broader context here is that Kedge challenges a common assumption in cloud computing. Most serverless platforms are designed to handle requests without remembering anything between them, and they send data to a separate managed database when it needs to be saved. Kedge makes data storage a core feature of the platform itself, copied across regions and accessible through SQLite rather than a proprietary tool. The 3-millisecond startup time, if it holds up under real-world traffic, addresses the cold-start problem — the delay users experience when a serverless app has to spin up from scratch — which has frustrated developers since AWS Lambda popularized the serverless model.

SQLite's use as the backbone for both the coordination layer and the user-facing database is also notable. SQLite was originally designed for single-device use, but has been moving into server and distributed contexts thanks to improvements in how it handles concurrent access and data replication. Kedge is betting that SQLite, with its simpler and more forgiving approach to keeping data consistent, is good enough for a global platform, avoiding the complexity of systems that require strict agreement between servers. Whether that bet works at large scale, particularly when network connections between regions are interrupted, is an open question.

The memory-sharing approach has precedent in other systems like Firecracker and Crosvm, though Kedge's specific combination of techniques is its own design. The memory savings matter most when running many small, similar app instances, which fits the platform's focus on fine-grained serverless functions.

In this author's view, Kedge's most interesting contribution is not any single component but how the pieces fit together: a globally shared database, simple web app building, near-instant startup, and per-second billing on actual usage, all in one platform. The 60-line Hacker News clone shows how much repetitive setup code a tight integration can eliminate. The open question is whether the simpler data-consistency approach can meet the demands of large-scale production use. For now, the platform is early, the billing model is freshly concrete, and the design choices are clear enough to evaluate on their merits.