What is celld from the NodeJS/Deno creator?

Quick check — I want to make sure I search for the right thing rather than guess, since this sounds like it could be a recent/niche project name.celld is Ryan Dahl's (creator of Node.js and Deno) latest project — a self-hosted, distributed reimplementation of Cloudflare's Workers and Durable Objects.

The core idea: Durable Objects are single-threaded, stateful serverless compute units, each with a globally unique ID and its own SQLite database, co-locating data and compute so you skip the usual distributed-systems headache of keeping cache and source-of-truth in sync. Normally that pattern only works if you're all-in on Cloudflare's infra. celld breaks that lock-in — it reimplements that exact API, but swaps Cloudflare's proprietary backend for any S3-compatible object storage and the Tokio async runtime in Rust.

How it's built (per Dahl's announcement): celld = V8 + S3 + SQLite + LTX + Tokio, programmed with exactly the Cloudflare Workers/DO JavaScript APIs and config, order of magnitude cheaper at scale, writes are durable before they're acknowledged with RPO=0, only depends on an S3 compatible bucket with no other infrastructure, and hibernated cells cost basically nothing beyond S3 storage.

Some concrete numbers (from daily.dev's writeup): it runs on your own VMs using V8 for JavaScript execution, one SQLite database per Durable Object (called a 'cell'), LTX for SQLite replication, and an S3-compatible bucket as both storage and coordination layer — eliminating the need for a separate control plane, Raft cluster, or consensus service, with roughly 4 MB RAM per resident cell, about 1,000 resident cells on an 8 GB node, ~4 ms cell wake time, ~90 ms durable write latency, and ~20 second failover.

Cost pitch: one comparison cited was running 100 continuously resident objects on Cloudflare would cost around $415 a month, versus roughly $49 a month for a celld deployment built on a DigitalOcean S3-compatible bucket.

It shipped as v0.1.0, Apache 2.0 licensed, alpha quality — one write-up put it plainly: it is also an alpha, so I would not move an important production system to it today, but I would absolutely build an experiment with it.

Given your interest in offbeat experiments with local/tiny models — the Durable Objects pattern (isolated compute + isolated SQLite per unit) is being talked up as a nice fit for spinning up per-agent sandboxes, which might be relevant if you ever want to run little Gemma/Qwen agent experiments with isolated state per instance.