ElectricSQL

Kevin De Porre presenting @electric-sql.com and @tanstack.com

Configurancy: Keeping Systems Intelligible When Agents Write All the Code | ElectricSQL AI makes code cheap; the scarce asset is the system's self-knowledge. Configurancy is the layer that lets bounded agents coherently co-evolve a system.

AI makes code cheap. The scarce asset is now the system's self-knowledge.

Specs and contracts were always valuable—but too expensive to maintain. So we invested sparingly, specs drifted, and we read the code instead.

That calculus has flipped with coding agents.

electric-sql.com/blog/2026/02...

Use the CLI to create & write to streams

4/Coming soon:

- Drop-in transports for Vercel AI SDK and TanStack AI
- Yjs support for collaborative editing
- HTTP proxy that makes existing SSE/token streams resumable with no code changes

3/ Agentic apps need a different coordination model.

Request/response assumes two parties taking turns.

Three agents and two users acting at once need a shared log everyone can read, resume, and react to.

That's durable streams.

2/Durable streams 0.2.0 ships with idempotent producers and exactly-once semantics.

400+ conformance tests. Client libraries in 10 languages: TypeScript, Python, Go, Rust, Java, Swift, PHP, Ruby, Elixir, .NET.

Announcing Hosted Durable Streams | ElectricSQL Hosted durable streams is now live on Electric Cloud—persistent, resumable event streams over HTTP with unlimited streams, reads, and writes.

electric-sql.com/blog/2026/01...

In 11 days from now, join the #LocalFirst and #CRDT devroom at #fosdem26 #fosdem in Brussels!
@electric-sql.com @tanstack.com

1/Hosted durable streams is live on Electric Cloud!

Persistent, resumable event streams over HTTP. Reads served from CDN, tested to 1M concurrent connections. 240K writes/second. 5M writes/month free.

🚀 Durable Streams 0.1.0 is live on npm

Two weeks since announcement → approaching 1,000 GitHub stars and people are already building with it.

Blog post — electric-sql.com/blog/2025/12...

Announcement blog post electric-sql.com/blog/2025/12...

GitHub - durable-streams/durable-streams: The open protocol for real-time sync to client applications The open protocol for real-time sync to client applications - durable-streams/durable-streams

github.com/durable-stre...

Today we're open-sourcing Durable Streams: a persistent stream primitive and HTTP protocol for reliable, resumable, real-time data streaming to clients.

We built it inside Electric. Now we're standardizing it as a standalone protocol.

TanStack DB 0.5 — Query-Driven Sync | TanStack Blog You don't need a new API for every component. With 0.5, the component's query is the API call. tsx // Your component's query... const { data: projectTodos } = useLiveQuery((q) = q .from({ todos }) .jo...

Blog post tanstack.com/blog/tanstac...

Designed for your existing REST, GraphQL, or tRPC APIs. No backend migration.

Bonus: Works with sync engines (@electric-sql.com, Trailbase, PowerSync) for real-time updates with near-zero incremental network cost.

🚀 TanStack DB 0.5 is here with Query-Driven Sync

Your component's query IS the API call. No custom endpoints. No GraphQL resolvers. Just write your query and DB figures out exactly what to fetch.

Details 🧵👇

The panel at Sync Conf 2025 brings together @tannerlinsley.com, @jamescowling.dev, @schickling.dev and @aaronboodman.com for a deep dive into challenges and tradeoffs between the technical approaches being explored in sync today.

Sync seems to be having a moment among web developers!

using bun and pglite from @electric-sql.com I'm able to create in memory or persisted postgres instances (one per bun process) available to use with Postgrex via `socket_dir`

in theory I could have a "cheap read replica" or simpler testing #myelixirstatus

One day boutique conference for the sync era of the web. Hear from the world's top builders about how to develop systems in the era of agentic software and real-time sync.

That's 168ms -> 1.6ms! and we have a lot of good ideas how to make it even faster.

The lesson: by shipping a system that works into production early and learning from it, allowed us to design a much better solution based on real-world requirements.

We've been running Electric in production for a year so we knew exactly what we needed: fast append-only writes and no data processing on reads.

The performance gains exceeded our expectations.

We're already syncing data to millions of clients in real-time behind a CDN, but our write path was heavily under-optimized.

The result: high CPU usage, accumulating WAL on Postgres, and high end-user p95 latencies.

Electric 1.1: new storage engine with 100x faster writes | ElectricSQL How we built a new storage engine for Electric, why we did it now, and how it delivers lower latency, higher throughput, and zero‑downtime deploys.

We've just delivered a 100x write performance bump with our latest release of Electric. How? We threw away our storage engine and built a new one designed for sync 🧵

electric-sql.com/blog/2025/08...

Bringing agents back down to earth | ElectricSQL Agentic AI, beneath all the hype, is actually just normal software. You can build agentic systems with a database, standard web tooling and real-time sync.

LangChain, vector databases, instruction routing, specialized memory stores.

You'd be forgiven for thinking you need a whole new stack to build agentic systems. However, that isn't actually the case: electric-sql.com/blog/2025/08...

How to build agentic systems on real-time sync.

We're just magic good sleep enablers! 😎

120 days of hardening – the post‑1.0 reliability sprint | ElectricSQL Our last quarter was a reliability sprint. We set one goal: make ElectricSQL so boring‑reliable that you stop thinking about it and just build.

electric-sql.com/blog/2025/08...

We fixed IPv6 fallbacks, WAL slot recovery, memory leaks, file handle exhaustion, replication races, and dozens more issues.

The boring stuff that makes infrastructure work. Because when AWS launched S3, they didn't lead with features — they led with eleven nines. 🎯