June 2025 Summaries

app.build is an open-source AI code generator initially focused on creating reliable CRUD web applications by prioritizing reliability over advanced features. The tool employs a finite state machine (FSM) for flow control, tree-based search for solution generation, and specialized actors for different tasks, ensuring that each component is validated before completion. Despite its initial limitations, app.build aims to enhance reliability through stateless design, concurrent execution, and a robust validation pipeline that includes TypeScript compilation and Playwright end-to-end tests. The tool emphasizes context engineering to avoid common pitfalls in AI code generation, utilizing encapsulation and feedback loops to refine its processes. Although initially constrained to a specific tech stack, app.build plans to expand its scope, allowing users to build more advanced applications with broader technological support, guided by insights gained from user interactions and error logs in its current iteration.

Postgres's default connection security, particularly the widely used `sslmode=require`, is criticized for providing insufficient security by failing to authenticate the server, making it vulnerable to man-in-the-middle (MITM) attacks. The article discusses efforts to improve this, including the introduction of `sslrootcert=system` in Postgres 16, which aims to enhance security by using system-trusted root certificates, although it's not yet widely supported across platforms, particularly on Windows. As a workaround, Neon has implemented channel binding via SCRAM-SHA-256-PLUS, which securely verifies server identity by requiring the server to prove it knows the user's password, thus protecting connections even without `sslrootcert=system` support. The article suggests potential future improvements, such as renaming `sslmode=require` to highlight its insecurity and creating a more secure default connection mode, indicating a hopeful trend towards improved connection security in Postgres.

Neon has introduced a native metrics export feature that allows users to stream Postgres metrics to any observability platform compatible with OpenTelemetry, expanding beyond the previous capability of streaming to Datadog. This enhancement addresses user demand for more comprehensive metrics integration, enabling users to view Postgres metrics alongside application and infrastructure data in a unified dashboard. Initially, users had to scrape their own metrics, but Neon now manages the scraping infrastructure using OpenTelemetry Collectors, which streamlines the process and reduces the need for user-managed infrastructure. The architecture involves deploying collectors per project, ensuring low-latency access to metrics and maintaining isolated failure domains. This setup allows for real-time monitoring and optimization of applications, providing a complete view of system performance. OpenTelemetry was chosen for its flexibility and ease of deployment, and it has been running reliably without requiring maintenance.

Neon has introduced the ability to export Postgres logs to any OpenTelemetry (OTLP)-compatible observability platform, addressing previously unmet user needs for direct access to log data alongside metrics. The company initially faced challenges in extending their existing metric export pipeline to logs due to limitations in the OpenTelemetry Collector's ability to scrape logs from Kubernetes pods by labels and the complexity of filtering log data from multiple processes running on NeonVM. To overcome these hurdles, Neon implemented a solution using rsyslogd to forward logs to the OpenTelemetry Collector in a structured format, enabling enrichment with Kubernetes metadata before exporting to platforms like Datadog. The implementation encountered obstacles, such as the inability to dynamically reconfigure rsyslogd without a restart and issues with matching logs to Kubernetes pods, necessitating additional steps to transform and enrich log data. The new logging capability not only enhances user experience by providing real-time access to logs but also supports compliance requirements, such as HIPAA, by enabling secure storage of Postgres logs.

Neon has been utilizing an AI tool named Devin to streamline their documentation processes, focusing on tasks like changelog setups, pull request (PR) reviews, and extensive edits across a large codebase. Devin is praised for its ability to autonomously handle repetitive and well-defined tasks, such as automating the setup of changelog files and addressing GitHub issues, which has significantly reduced manual labor. It also exhibits context awareness by integrating relevant images into documentation without explicit prompts. However, Devin's limitations are evident in complex, ambiguous tasks and its black-box nature, where it operates without interactive feedback, making it less reliable for tasks requiring nuanced judgment. Despite these challenges, Devin has been a valuable assistant for Neon, enabling their team to work more efficiently by offloading certain clerical tasks, although there is a desire for more interactive capabilities to enhance its usability.

Rubber duck debugging, a technique where developers explain their coding issues to an inanimate object to gain new insights, is evolving with the integration of AI in IDEs like Cursor. This conversational approach enhances debugging by allowing AI to act as an interactive partner, which can identify logical gaps, ask probing questions, and build a richer context around the problem than traditional AI methods. While simple error-checking can be efficient at times, relying solely on blindly applying AI-generated solutions may impede learning and understanding. By using a pre-defined ruleset and voice-to-text integration, developers can initiate consistent and productive debugging sessions that mimic the collaborative benefits of discussing issues with a colleague, thereby improving both understanding and solution accuracy.

Databutton, now rebranded as Riff, has enhanced its platform by integrating Neon, a serverless Postgres solution, to provide production-ready infrastructure with every app, eliminating the need for setup and external services like Supabase for data persistence and authentication. This integration allows the Databutton AI Agent to manage the entire app stack, including creating and managing Postgres databases, schema design, data migrations, and user authentication seamlessly, thereby reducing complexity for users and enhancing their building experience. Neon enables each app to have its own dedicated Postgres database, which can be provisioned in seconds, automatically suspended when idle, and resumed quickly, offering high efficiency and scalability without the typical cloud infrastructure costs. This upgrade empowers users to create full-stack web applications with zero coding effort, as the AI Agent handles everything from frontend and backend development to database management, making the platform accessible even to non-technical users.

DEV, a prominent platform for developers, transitioned from Heroku to Neon to better handle its growing database demands and enhance its developer-focused ethos. Initially relying on Heroku Postgres, DEV encountered limitations related to scalability, cost inefficiencies, and operational overhead. These challenges prompted a search for a more adaptive solution, leading to Neon, which offers serverless Postgres with autoscaling capabilities, eliminating the need for overprovisioning resources. Neon's developer-centric approach, intuitive user interface, and API-first design resonated with DEV, allowing for seamless automation and integration with tools like Vercel. Additionally, Neon's lightweight read replicas and branching model provided significant advantages over traditional setups, enabling DEV to efficiently manage testing environments and reduce maintenance costs. This shift not only addressed technical needs but also fostered a collaborative partnership between DEV and Neon, emphasizing a shared commitment to providing developers with modern, efficient infrastructure solutions.

The MCP server's testing process is designed to ensure that large language models (LLMs) can effectively select and utilize the appropriate tools from a suite of over 20 available options, including unique tools for database migrations. The server includes "prepare_database_migration" and "complete_database_migration" tools that facilitate SQL-based database migrations by creating and applying changes on temporary and main branches, respectively. To address the complexity LLMs face in maintaining the correct workflow, the team implemented evaluation tests, or "evals," using a method called "LLM-as-a-judge" to ensure that LLMs use the tools in the correct sequence. These evaluations, which include the "factualityAnthropic" and "mainBranchIntegrityCheck" components, help verify the factual accuracy and integrity of the database migrations. By refining the descriptions for these tools, the team improved their pass rate from 60% to 100% without additional coding. The use of the Braintrust platform provides a user-friendly interface for monitoring and analyzing test runs, underscoring the importance of rigorous testing and the benefits of managed services in software development.

Neon has launched a Startup Program offering up to $100,000 in credits for early-stage companies to build on its platform, aiming to alleviate concerns around database scaling, security, and costs. The program, open to startups with $1M to $5M in venture funding, provides access to Neon's 100% Postgres database known for its speed, scalability, and features like instant branching and seamless AI-native development. Participants benefit from early access to new features, priority support, and visibility opportunities, while the program highlights Neon's commitment to supporting startups from inception by providing a database infrastructure that adapts automatically to varying demands.

Neon provides significant advantages in compute efficiency and cost savings compared to provisioned platforms, utilizing autoscaling to optimize resource allocation dynamically. Production databases on Neon require 2.4 times less compute and incur 50% lower costs than those on provisioned platforms like AWS RDS, due to the ability to adjust resources in real-time, avoiding over-provisioning and under-utilization issues. Neon read replicas are 4 times more efficient, and the autoscaling feature allows small, non-production databases to scale to zero, yielding even more dramatic savings. This approach minimizes performance degradation incidents and reduces the operational burden on users, as evidenced by a real-world example where a Neon database incurs 60% lower costs than its provisioned counterpart. The autoscaling mechanism adjusts resources approximately every 81 seconds, ensuring optimal performance and cost-effectiveness, even during load spikes. Overall, Neon's autoscaling capability provides a flexible, efficient, and economical solution for managing database workloads, challenging the traditional provisioned model.

Implementing full-text search in PostgreSQL using basic ILIKE queries can be straightforward for small datasets but becomes inefficient at scale due to the lack of effective indexing, prompting the need for advanced solutions like PostgreSQL's native full-text search (FTS) or ElasticSearch. Native FTS with PostgreSQL involves using tsvector and tsquery data types along with GIN indexing, which works well for small to medium datasets but struggles with performance and search quality as datasets grow, lacking features like fuzzy matching and efficient result ranking. ElasticSearch, on the other hand, excels in performance and offers extensive search features but introduces significant operational complexities due to the need for data duplication, synchronization, and additional infrastructure management. pg_search emerges as a compelling alternative, embedding modern search capabilities directly within PostgreSQL through a Rust-based extension, offering features such as BM25 ranking, fuzzy search, real-time indexing, and reduced operational overhead compared to maintaining a separate ElasticSearch cluster. By integrating these advanced search functionalities into PostgreSQL, pg_search provides a balanced solution that delivers ElasticSearch-quality search performance with the simplicity and consistency of a single-system architecture, making it an attractive choice for teams already utilizing PostgreSQL and desiring enhanced search capabilities without additional complexity.

Handling Protected Health Information (PHI) under HIPAA regulations is crucial for developers building healthtech applications, as violations can lead to significant financial penalties and a loss of patient trust. PHI encompasses any health data that can identify an individual, and developers must ensure compliance through encryption, access controls, and secure APIs. De-identification methods, such as the HIPAA Safe Harbor and Expert Determination, help transform PHI into non-PHI data while maintaining data utility. Developers should avoid using real PHI in development environments, ensure secure storage and access controls, and implement rigorous logging and monitoring practices to detect unauthorized access. Building a security-first culture within healthcare technology is essential to maintaining patient trust and complying with HIPAA regulations.

Microsoft has introduced a new Visual Studio Code extension for managing Postgres databases, which integrates GitHub Copilot to enhance user interaction with databases directly within the editor. This extension provides functionalities similar to traditional tools like PgAdmin or DBeaver, including schema visualization, database connection, and integration with Azure for managing Postgres instances. It features a "Chat with Database" capability, allowing Copilot to interact with the database schema for insights, as demonstrated with Sentry's complex Postgres schema. Additionally, the extension facilitates generating SQL migration scripts and offers IntelliSense for SQL query writing, enhancing productivity for developers working with sophisticated database environments. The integration aims to provide a seamless development experience by embedding database management tools directly into the coding environment, exemplified by exploring Sentry's schema and understanding its structure and relationships.

Neon experienced two significant outages in May 2025 due to IP allocation failures in their AWS us-east-1 region, primarily caused by a misconfiguration of the AWS CNI plugin that allocates networking resources for Kubernetes Pods. The disruption began when a periodic job in Neon's Control Plane failed to terminate idle databases, leading to an unexpected surge in active Compute Pods and exhausting available IP addresses in the VPC subnets. Despite initial configuration changes intended to free up IPs, the issue persisted due to the AWS CNI plugin's default behavior of holding IPs in a cooldown state and inefficient allocation practices. Subsequent attempts to stabilize the system by modifying the WARM_IP_TARGET setting inadvertently worsened the situation, as the configuration led to IP allocation errors that prevented new Pods from starting. The investigation revealed that the AWS CNI plugin's handling of IPs was inconsistent with expectations, particularly when WARM_IP_TARGET was set to values that conflicted with the underlying IP management logic. Eventually, the errors subsided due to a temporary halt in idle compute terminations, which allowed for IP reallocation. The team at Neon has since submitted a pull request to AWS CNI to address these issues and shared their findings publicly to aid other teams in avoiding similar problems.

Neon Launchpad is a tool designed to simplify the process of integrating Postgres databases into development projects by allowing users to generate a hosted Postgres database instantly without requiring setup, login, or configuration. This tool provides a ready-to-use connection string compatible with any Postgres-supported framework and offers the flexibility for users to claim their database later via a Neon account if desired. Databases created with Launchpad last for 72 hours by default, and the tool is particularly beneficial for embedding per-user Postgres databases into developer platforms, AI agents, and open-source projects. Neon has open-sourced the key packages behind Launchpad, including a CLI helper and a Vite plugin, to make it more extensible for various projects. Future enhancements planned for Launchpad include features like schema pushing, data seeding, and client bootstrapping, with developers encouraged to provide feedback and contribute to its evolution.

Netlify has introduced Netlify DB, a production-ready Postgres database powered by Neon, as part of its efforts to become a comprehensive platform for AI-native development. This new feature allows developers and AI agents to provision databases quickly and easily, offering instant access to infrastructure without the need for external signups. The integration supports the concept of Agent Experience (AX), emphasizing the importance of providing a seamless environment for both human developers and AI agents by ensuring databases are fast, scalable, and accessible. Neon enhances this offering by enabling rapid provisioning, automatic scaling to zero, and API-first management while maintaining compatibility with standard SQL and popular Postgres extensions. Users can initiate Netlify DB through the Netlify CLI or Dashboard, seamlessly integrating it into their workflow with minimal setup, and can later link it to their Neon account for enhanced management.

app.build is an open-source AI agent designed to create and deploy full-stack applications, offering features such as end-to-end tests and automated deployments. Despite discontinuing its managed version, the source code remains available for those interested in building similar agents. The tool allows users to generate applications from scratch or iterate on existing ones, and it supports local execution, enabling developers to use their preferred models. By employing a divide-and-conquer approach, app.build ensures high-quality applications by breaking down app creation into independent tasks that undergo rigorous checks. The project aims to foster a community of developers to contribute and innovate, with plans for a public roadmap and further resources to support users in understanding the agent's architecture and capabilities.

SMASHSEND, an AI-driven email marketing platform, drastically reduced its database costs by transitioning from AWS RDS to Neon, achieving a 95% cost reduction and solving several operational challenges. Initially, SMASHSEND faced issues with RDS due to its inability to efficiently handle bursty workloads, high connection demands causing failures in background jobs, and expensive replication for testing environments. The migration to Neon was completed in about three hours, involving snapshot import, environment variable updates, and gradual traffic rollout, leading to significant improvements. Neon’s serverless architecture allowed for automatic scaling and cost reductions by adjusting resources based on activity levels. Built-in connection pooling resolved previous job failures, and the introduction of ephemeral Neon branches facilitated more efficient testing and integration processes, aligning seamlessly with SMASHSEND's serverless infrastructure. The transition not only improved cost efficiency but also enhanced performance and workflow integration, making Neon a compelling choice for startups facing similar challenges with RDS.

Neon's engineering team has been leveraging Cursor to enhance their coding efficiency across languages such as TypeScript, Go, Rust, and Python, finding particular success with features like Cursor Tab for context-aware suggestions and Agent Mode for scaffolding and debugging. Cursor has proven helpful in understanding unfamiliar code and modernizing legacy systems, while its swift adoption of new models is appreciated. However, limitations exist, including issues with context awareness, incorrect suggestions, and reliability challenges in large files. The team suggests that Cursor could become indispensable with improved context sensitivity, integration with project-specific details, and a more comprehensive feedback loop. Despite its shortcomings, Cursor has already delivered value by reducing ramp-up time and accelerating repetitive tasks, although the team is not yet using it for end-to-end production code. Neon, a serverless Postgres database, effectively supports AI IDEs like Cursor, emphasizing the tool's potential when integrated into their workflows.