May 31, 2026
S3, tea, and database drama
Show HN: Streambed – Stream Postgres to Iceberg on S3, Supports Postgres Wire
A new tool says it can spare your main database — but commenters are already arguing over the fine print
TLDR: Streambed is a new tool that copies changes from a live Postgres database into cheap cloud storage so people can run reports without slowing the main app. Commenters liked the idea but immediately fought over the bold "no ETL" claim, with skeptics saying real analytics is never that simple.
A shiny new Show HN launch promised something irresistible to overworked engineering teams: copy data changes out of a busy Postgres database, store them cheaply on S3, and let people run heavy reports somewhere else — without changing the app. In plain English, it’s pitching itself as a way to stop analytics dashboards from bullying your production database. The creator says it was born from a very real workplace headache at Cloudflare, where the usual answer was apparently "ugh, guess we need another custom setup again."
But the comments? Instant reality check. The biggest eyebrow-raise came from users pushing back on the slogan-like promise of "No ETL" — basically, "no complicated data-moving pipeline." One commenter flat-out said that’s not really true, because once a company gets serious about analytics, they’ll still need to clean up and reshape the data at some point. Translation: the crowd was not about to let marketing have a free pass. It turned into a classic Hacker News scene: one side saying, "This is genuinely useful," the other saying, "Cool demo, but let’s not pretend this magically solves everything."
There was also some professional respect in the thread. A builder of a rival-ish tool popped in with a very polite "we’re in the same space" nod, which is basically startup-comment-section diplomacy at its finest. And then came the most practical crowd energy of all: show us the numbers. Because no matter how elegant the pitch sounds, the internet wants receipts, benchmarks, and proof that this thing can actually keep up when real workloads start throwing punches.
Key Points
- •Streambed is a CDC engine that streams PostgreSQL WAL changes to Parquet files on S3 and commits Apache Iceberg metadata.
- •The system is intended to offload analytical queries from a production Postgres database without requiring application changes.
- •Streambed can expose Iceberg tables through a built-in query server that uses embedded DuckDB and the Postgres wire protocol.
- •Its main data flow is logical replication from Postgres, row buffering by table, periodic Parquet flushes to S3, and Iceberg commits, with copy-on-write merging for updates and deletes.
- •The project includes commands for continuous sync, one-shot resync, standalone querying, and cleanup, and it requires Go 1.22+ and CGO for development.