Project
Apache Parquet is an open-source columnar storage file format optimized for compression and query efficiency, widely used in big data and analytics systems.
Iceberg 1.11.0 ships a File Format API that sends Spark, Flink, Hive and plain-Java reads and writes through one registry of format models. A dev.to walkthrough says the file format list stays fixed, so Vortex and Lance still need a change inside Iceberg.
Reality
Infino, from OpenSearch veterans, launched a platform letting agents query one Parquet copy in object storage in place of four separate data systems. The pitch is fewer pipelines and one place to set access policy, and so far it rests on the CEO's account of how agents query.
Reality
QuerySurge 15 links data-mapping documents to the validation tests its AI generates from them and graphs which tests a mapping change affects. The company has not said how accurate those tests are, so full coverage can still include wrong tests.
Reality
A dev.to walkthrough found DuckDB's operator names in the EXPLAIN output of AWS's aurora_analytics extension, on a Parquet query that executed in 81 ms. That run hit a warm cache on one 500,000-row file, so it tells you which engine ran far more than how fast it will be on lake-scale data.
Reality
AWS has embedded DuckDB in Aurora PostgreSQL from versions 17.11 and 18.6, so one SQL query can join live rows with Iceberg and Parquet tables in S3. Reverse-ETL jobs that exist only for that join can be retired if the database instance has room for the lake scans.
Reality
AWS Community Builder Aki argues Apache Iceberg made data portable while permissions, auditing, lineage and sharing stay tied to each catalog. The spec leaves the atomic commit to each catalog, so the catalog is now the choice that is hard to reverse.
Reality
Reviewers logged 10,991 passing tests on Iceberg's first 1.12.0 candidate before rejecting it over two vendored web files. The change most likely to break a downstream build is a removed partition-statistics helper.
Reality
Keeping the Iceberg table format governance-unaware was a deliberate call. That leaves the catalog as the only thing standing on the path of every read. A dev.to guide lists six pieces an operator has to source above it.
Reality
A one-line change to the open call leaves a real database file in the browser, with a write-ahead log and checkpoints that survive a restart. Getting it to persist depends on which DuckDB-Wasm build you installed.
Reality
A dev.to post argues that Snowflake-managed Iceberg tables let you keep the managed engine while owning the Parquet in your own bucket. The same post says the catalog stays Snowflake's, and that 50,000-inserts-per-second ingest should not move.
Reality
A GCP walkthrough put Spark, Trino and BigQuery on the same orders table through Google's Iceberg REST catalog. The 409 ConflictException it logged is the part of the demo a production pipeline has to plan for.
Reality
KeewanoDB keeps events in the order they arrived, grouped around a customer or device, so an agent can read one entity's history without an ETL job. Keewano expects the warehouse to stay where it is.
Reality
A dev.to walkthrough puts a quantized Llama 3.2 3B on up to 10,000 concurrent Lambdas and writes a million daily briefings for $48. The timing math is internally consistent. The per-GB-second price behind the $48 is unstated.
Reality
Microsoft documents no configurable mirroring window, and support confirmed that restarting a mirror reloads the whole table instead of resuming from the last CDC position. The estate here has roughly 2,500 tables.
Reality
Uber reported 169:1 on 200TB of Spark logs a day by separating static templates from variable fields before anything was written to HDFS. Whether that ratio transfers depends on how few call sites your lines come from.
Reality
The figure comes from a conference demo scanning a 4K video, and the same talk sets a target closer to 50 Gbit/s. The reason today's loaders still stage shards on disk turns out to be PyTorch's worker model.
Reality
A distinct-value sketch is thousands of bytes and a delete bitmap has no fixed size, which is why Iceberg keeps both in a separate file, and why a bad join plan is a statistics problem rather than a rewriting one.
Reality
Native geometry and geography types in format v3, with matching Parquet 2.11 logical types underneath, give a shape column its own bounding boxes and a CRS that travels with the data. The second half is the part worth migrating for.
Reality
Hebbian Robotics, a two-person YC Summer 2026 company, announced HFlow on August 19 and shipped it to PyPI the next day, betting that robot dataset checks and provenance are worth sharing between teams rather than rebuilding.
Reality
Lock-in migrated from the model to the runtime that holds your tool grants, approval rules and learned state. Six specifications aim to make that runtime's contents portable, and half of them are one person's drafts.
Reality