EDB Engineering Newsletter #16

Mar 02, 2026

Issue #16 is here. From the drawing board to the codebase, here’s a look at what the EDB Engineering team has been up to this month.

News we’re watching:

ggml.ai joins Hugging Face

Georgi Gerganov’s ggml.ai has joined Hugging Face to accelerate their joint mission of making local AI easy and efficient for everyone to run on their own hardware. Currently, when a new model (like Llama 3 or Gemma) is released, the community has to manually port it to llama.cpp. With this merger, that wait time is expected to shrink from days to hours as the transformers library becomes directly compatible with the ggml ecosystem.

Writing a good CLAUDE.md

Research into improving the implementation speed and accuracy of coding agents has revealed that popular automated onboarding methods may be counterproductive. A recent study from ETH Zurich show that LLM-generated context files actually reduce task success rates by approximately 2% to 3% compared to providing no repository context at all, while simultaneously increasing inference costs by over 20%.

For data infrastructure and engineering teams, this highlights a critical need for high-leverage “context engineering”: instead of relying on token-heavy automated summaries, developers should manually craft concise CLAUDE.md or AGENTS.md files that utilize Progressive Disclosure to point agents toward specific architectural docs only when needed.

From the EDB team:

Why AI accountability cannot be outsourced

As AI systems become more autonomous, we are forced to confront a foundational question: When the machine makes the call, who holds the liability? EDB Chief Legal Officer Rob Feldman shares his thoughts on the blog.

Why OpenAI should use Postgres Distributed

Database Engineering lead Jozef de Vries tackles critique of Postgres’s scalability on the blog. Following OpenAI’s deep dive into supporting 800 million users, MariaDB suggested a total platform switch was necessary.

Jozef argues otherwise: the strength of Postgres lies in its vibrant ecosystem and innovation. With EDB Postgres Distributed, organizations can handle modern, high-scale demands through a distributed architecture while sticking with the Postgres ecosystem.

Claiming Operational Independence for PostgreSQL

Sovereignty over the data layer is the mandatory first step for operational independence and Sovereign AI. If you already run applications in a Kubernetes-managed service provided by a hyperscaler but rely on their proprietary DBaaS for your database, you can reclaim sovereignty by moving the database into that same Kubernetes cluster with CloudNativePG and PostgreSQL.

Gabriele Bartolini shares how in his post, and looks ahead to his KubeCon session on the topic.

We made it onto the CNA Enrichment Recognition List

An incredibly rewarding moment as EDB has been named to the CNA Enrichment Recognition List. Out of thousands of global organizations, we’re among a select group recognized for consistently providing gold standard vulnerability data. As a recognized CVE Numbering Authority, EDB publishes CVE records with CVSS severity and CWE classification at the time they’re released.

A stack-buffer-overflow exercise with AddressSanitizer and PostgreSQL

Mark Wong (Senior Software Developer and PostgreSQL Major Contributor) dives into the world of memory safety. Read on for a story of how AddressSanitizer successfully identified a memory corruption bug, along with a “How-To” for getting the tool up and running with your local PostgreSQL regression tests.

Compiling PostgreSQL extensions with Visual Studio 2026 on Windows

Xavier Fischer, a .NET Principal Software Engineer on the connectors team, recently took a deep dive into a common user pain point: compiling PostgreSQL extensions within Visual Studio. The result is a practical, step-by-step tutorial on how to get things running smoothly using Visual Studio 2026 Community.

Dynamic Extensions in Kubernetes with PostgreSQL 18

Thanks to new capabilities in PostgreSQL 18 and Kubernetes 1.33, you can now mount and load extensions like pgvector as read-only volumes using OCI images. Ramalingam Srinivasan explains how this improves security and agility for EDB Postgres AI users.

Growing the PostgreSQL community team

In 2025 we started a program (“Developer U”) to help colleagues who show promise for PostgreSQL Development to become contributors. At the start of February this group met for another on-site. Read more to learn about their experience this time around: https://www.enterprisedb.com/blog/developer-u-exercising-cohesion-and-technical-skill-postgresql

Until next time

We hope you enjoyed this edition of the EDB Engineering Newsletter! Closing out with our Slack highlight of this month, Artjoms Iskovs spotted this gem of an error message:

Until next time,

The EDB Engineering Team

EDB Engineering Newsletter