September 2025 Summaries

DAT Systems, a Pakistan-based startup, efficiently manages field staff operations for over 16,000 workers at the Lahore Waste Management Company by leveraging Neon, a serverless Postgres provider. Using messaging platforms like WhatsApp, DAT's system processes around 10 million transactions monthly, with each message triggering multiple background jobs and audit logs to ensure accountability. Neon’s serverless features, such as autoscaling, read replicas, and scale-to-zero, enable DAT to handle high transaction volumes without downtime and optimize resource usage during peak and idle periods, respectively. Additionally, DAT utilizes Retool to quickly deploy customized real-time dashboards for clients, enhancing operational oversight with maps, tables, and forms. Neon’s branching feature facilitates efficient development and testing environments, allowing DAT to innovate without incurring excessive costs.

Implementing a Kalman Filter directly within Postgres offers a novel approach to smoothing noisy GPS data, traditionally handled outside the database in environments like Python or MATLAB. This method leverages SQL functions, recursive queries, and aggregates to refine GPS data by addressing the inherent noise from satellites, buildings, and devices, making large-scale analysis more reliable. The Kalman Filter is a recursive algorithm that estimates the true state of a dynamic system by combining predictions with observed measurements while explicitly modeling uncertainty. Challenges in implementing it within SQL include maintaining state, updating it with each new measurement, and ensuring data is processed in sequence. The project by traconiq includes an open-source implementation that supports both online and offline filtering, with benchmarks indicating that while online filtering ensures immediate availability of smoothed data, offline filtering, especially using custom aggregates, offers higher efficiency for large datasets. This approach significantly enhances the ability to perform scalable analytics directly within Postgres, offering immediate, realistic paths from jittery GPS samples useful for analytics or visualization.

Postgres 18 has been released with a variety of enhancements aimed at improving performance, flexibility, and the developer experience, and is immediately available for preview on the Neon platform. Key features include B-tree skip scans for more efficient query processing, improved EXPLAIN output for better query optimization, and virtual generated columns that don't increase table size. The update also introduces native UUIDv7 support for improved index performance, an expanded RETURNING clause for capturing old and new row values, and temporal constraints for maintaining integrity in time-based data. Additionally, Postgres 18 supports OAuth authentication and introduces asynchronous I/O for read operations, enhancing throughput by allowing multiple reads to occur simultaneously. The release also brings refined autovacuum maintenance configuration and parallel GIN index builds to boost indexing performance. While Neon is currently optimizing its backend to fully integrate these new features, users are encouraged to try Postgres 18 on Neon's Free plan, with full production readiness expected once the preview phase concludes.

Elephantshark, an open-source Ruby script developed by Neon, is designed to monitor and troubleshoot Postgres network traffic by acting as a middleman in Postgres-protocol exchanges. Unlike Wireshark, which struggles with SSL/TLS-encrypted connections, Elephantshark can decrypt and re-encrypt Postgres connections, parsing and logging messages between servers, clients, drivers, ORMs, proxies, and poolers. It facilitates detailed monitoring by forwarding and annotating network traffic, and it can also work alongside Wireshark by writing keys to an SSLKEYLOGFILE. The tool demonstrates its functionality through a step-by-step process of connecting to a Neon Postgres database, showing the detailed exchange of bytes between the client and server, including authentication and parameter statuses. Users can access more information and installation instructions via the README on GitHub.

The online_advisor is a Postgres extension designed to optimize database performance by analyzing real-time query workloads and providing actionable recommendations for indexes, extended statistics, and prepared statements. By integrating with Postgres’ executor, it monitors queries to suggest potential performance improvements such as creating indexes when queries filter numerous rows, proposing extended statistics when planner estimates are inaccurate, and identifying queries that would benefit from prepared statements to reduce planning overhead. While it does not automatically implement these suggestions, users can review and apply the recommendations to enhance query efficiency. The extension is compatible with Postgres versions 14–17 and is available on the Neon platform specifically for Postgres 17, offering significant benefits for managing complex query environments by addressing issues like misestimated row counts and unoptimized query planning.

Neon has introduced a new feature that allows developers to create snapshots via an API, enabling versioning and checkpoints for agentic platforms. This feature is built on Neon's copy-on-write branching architecture, which allows users to jump between different app versions while preserving not only the code but also the exact database schema and data. This capability is particularly beneficial for platforms where code and database changes occur simultaneously, as it helps avoid issues like broken queries and mismatched data. Neon’s architecture, which separates compute from storage, supports this functionality by enabling efficient storage of database changes as deltas, allowing for quick creation and restoration of snapshots without the need for full data duplication. Additionally, Neon offers an API-first approach for provisioning and control, making it easy to integrate this checkpointing system into an agent's workflow without requiring a user interface. The initiative is designed to enhance user experience by providing a seamless way to manage and revert app changes, making the platform an attractive choice for developers building full-stack AI agents.

Marble, an open-source headless CMS, has improved its shipping speed by utilizing Neon branching, an innovative database feature that allows testing of big migrations safely with production data. By adopting Neon as their primary database, Marble benefits from its simplicity and efficiency, enabling quick project setup and seamless collaboration. Neon branching offers Marble the ability to create instant branches from production data, facilitating rapid testing and development without the risk of impacting live environments. This approach allows Marble's small team to maintain high shipping velocity by testing code changes against real production data, thus ensuring the reliability and accuracy of new features, particularly for complex updates like analytics dashboards. Marble's tech stack leverages a serverless architecture with tools like Astro, Tailwind CSS, Next.js, and Cloudflare Workers, while Neon, integrated with Prisma, serves as the main database. The Neon Open Source Program supports projects like Marble by increasing resource limits, fostering growth, and enabling swift and secure feature deployment.

OpenAI has introduced support for MCP servers in ChatGPT, enabling users to connect external tools such as Neon's MCP server to manage Postgres projects directly from the chat interface. This integration allows users to query, monitor, and manage database activities by transforming ChatGPT into a comprehensive command center for infrastructure and applications. The MCP protocol facilitates seamless communication between AI editors and external services without the need for custom connectors for each integration. While this feature enhances productivity by allowing real-time interaction with databases, it is recommended for local development and IDE integrations rather than production environments due to potential security risks. The setup involves copying the MCP server URL, creating a connector in ChatGPT, authorizing through OAuth, and approving access in Neon, after which users can efficiently manage Neon projects using ChatGPT.

Neon Testing is an open-source utility developed by Mikael Lirbank that simplifies database integration testing by automating the lifecycle of Neon branches for tests conducted with Vitest. It provides isolated, disposable test environments by creating a fresh Postgres branch for each test file, ensuring that tests are run against a realistic copy of the production database schema with all constraints intact. This approach addresses the challenges of managing containers, seeding data, and applying migrations typically associated with integration testing, thereby reducing the risk of false positives and facilitating the identification of issues that might arise in production. Neon Testing integrates seamlessly with Vitest, allowing developers to focus on writing tests rather than managing infrastructure, and is available for use via npm and GitHub.

Neon has shifted away from traditional flat-fee structures for premium features, instead adopting a usage-based pricing model to ensure fairness and scalability for their customers. This change allows essential features, such as HIPAA compliance and SOC 2, to be accessible to all users at a cost proportional to their usage rather than being locked behind expensive enterprise plans. Inspired by the serverless pricing model of Databricks, Neon's approach eliminates arbitrary plan gating, allowing smaller companies to avoid subsidizing larger enterprises. By introducing a higher compute unit price for accessing premium features and applying usage-based surcharges for especially costly features, Neon ensures that customers only pay for what they actually use, providing a more equitable solution for companies of all sizes.

Neon's Free Plan provides developers with a robust platform to start building applications using Postgres, allowing for up to 20 projects, each with its own resource limits, such as 0.5 GB of database storage and 100 CU-hours of compute time per month. To maximize the utilization of these resources, users are advised to monitor each project separately, optimize compute usage by scaling down CPU allocations, and practice storage housekeeping by managing data growth through retention policies and cleaning up unused indexes and dead tuples. Additionally, developers should be mindful of network transfer limits by optimizing queries to avoid unnecessary data retrieval and manage ORM defaults. By adopting these practices, developers can effectively extend their use of the Free Plan, making it a versatile tool for prototyping, testing, or launching Postgres-powered applications.

AI coding assistants like Cursor can be enhanced with Neon-specific context through a set of AI rules that guide them in generating more accurate code for Neon applications. These rules, contained in small .mdc files, help the AI understand how to handle specific contexts in a codebase by defining applicable file types and offering descriptions of their use. For Neon, these rules provide tailored guidance for best practices, particularly in areas like authentication, serverless connections, and integration with Drizzle ORM. For example, the neon-auth.mdc rule provides the correct patterns for setting up authentication in Neon, while the neon-serverless.mdc rule ensures proper handling of serverless connections to prevent common mistakes. Similarly, the neon-drizzle.mdc rule optimizes the setup of Drizzle ORM with Neon, ensuring efficient schema design and query optimization. These AI rules can be easily integrated into projects, allowing developers to leverage Neon’s features effectively for building scalable applications.