November 2025 Summaries

For Thanksgiving, the author reflects on their experience of building an app using Google's AI-native IDE, Antigravity, and the open-source platform Harper, despite having no coding background. The project involved directing AI tools like ChatGPT to translate ideas into requirements and using models such as Gemini 3 and Claude Sonnet 4.5 for development. The result was a Thanksgiving-themed game where players navigate a maze, avoiding turkey emojis and collecting food across five levels. The app tracks player activity with CRDT counters and features a leaderboard that updates frequently. The author highlights the ease of deploying the app on Harper Fabric and the newfound accessibility of turning ideas into reality with AI tools, suggesting a shift toward a more imagination-driven era of software development.

Ivan Judson, a Solution Architect at Harper, elaborates on the creation of a data integration solution utilizing Harper and BigQuery, focusing on the current data flow and the development of a synthetic data generator. He outlines the integration process with BigQuery and the configuration of the data synchronization process, while also discussing prospective enhancements aimed at improving the project. Additionally, Ivan underscores the importance of community engagement and collaboration in future developments.

Austin's podcast features an interview with Colin and Josh, the creators of MELL, an AI-powered community reporting app designed to simplify the process of addressing local issues such as potholes and broken streetlights. During the discussion, the developers delve into their ideation process, technical backgrounds, and the challenges they faced while building the app, while also sharing their vision for future enhancements. The conversation underscores the significance of user experience and the app's potential to foster meaningful change within communities, highlighting its capacity to improve local infrastructure reporting and response.

Harper presents a transformative approach to real-time data architecture, essential for AI systems that require rapid decision-making and seamless operation across varied environments, from cloud to edge. Traditional architectures, which rely on a patchwork of MQTT brokers, message queues, and databases, introduce latency and complexity, particularly in unstable network conditions. Harper addresses these challenges by integrating pub/sub, MQTT, WebSockets, SSE, storage, replication, and vector search directly within the database, creating a unified platform for AI. This architecture allows for low-latency event processing, durable state feeding models, and reliable replication in 50–100 milliseconds. It supports multiple protocols and ensures consistent behavior regardless of deployment environment, thereby bypassing the risks associated with cloud service deprecations and network inconsistencies. Harper’s design simplifies operational complexity, offering a reliable, scalable, and portable solution that aligns with the demands of modern AI applications, enabling organizations to build responsive, intelligent systems without the burden of conventional infrastructure.

Software architecture has undergone phases of complexity and simplicity, evolving from monolithic systems to microservices and now to distributed microliths, which aim to reduce complexity while maintaining scalability. Unlike traditional architectures that scale by adding more systems, the distributed microlith achieves efficiency by replicating structure within a unified runtime, allowing modular components to interact directly without network-induced latency. This approach is exemplified by platforms like Harper, which demonstrate significant performance improvements in real-world applications such as e-commerce and real-time data processing, by eliminating inter-service communication overhead and maintaining a cohesive system across regions and clouds. The distributed microlith model emphasizes simplicity as a sustainable method for scaling, proving that cohesive systems can achieve high performance and resilience without the need for extensive infrastructure, complex orchestration, or fragmentation.

In a Harper Learn session, Solutions Architect Ivan Judson demonstrates the process of transforming a simple local setup into a fully managed, multi-node Harper cluster with ease. He showcases the seamless transfer of configuration, access groups, and application logic from a laptop to the cloud, emphasizing Harper's streamlined deployment workflow that enables immediate scaling across nodes. The session illustrates how Harper's platform facilitates the swift transition from local development to distributed production environments. Ivan also previews the next session by sending an MQTT message to trigger a real-world action, highlighting Harper's capability in supporting fast, event-driven applications.

Harper is a unified development platform designed for enterprises requiring fast and resilient backends for both user and bot experiences, significantly enhancing performance by moving applications closer to the edge. It integrates database, cache, application, and messaging into a single high-performance runtime, offering prerendering capabilities that improve SEO visibility and infrastructure resilience. In contrast, Prerender.io serves as middleware focused on serving cached HTML snapshots to crawlers, enhancing crawlability and indexation without improving human user performance. Harper's edge-distributed architecture supports both bot and user delivery, maintaining uptime and performance even when origin systems fail, while offering real-time data freshness and scalability across various workloads. Prerender.io, however, relies on external uptime and operates solely at the SEO middleware layer, making it less comprehensive for enterprises seeking full-stack performance gains. Harper's platform value increases with scale, providing a robust solution for dynamic data and large catalogs, whereas Prerender.io remains a single-function tool limited to prerendering.

Harper Hack 2025, a two-week global hackathon, celebrated the open-source launch of Harper, inviting developers to explore its unified development platform. The hackathon was notable for the impressive quality of projects created, showcasing the platform’s capabilities in innovative ways, such as AI-driven interfaces and civic engagement tools. A live kickoff event introduced judging criteria emphasizing technical excellence and thoughtful design, encouraging developers to leverage Harper's open architecture for rapid development and AI integration. The community engaged actively through discussions and collaborations, culminating in three standout projects: AI Gesture Control, ConvoIQ, and Mell, with Mell winning the grand prize for its civic-minded application using Harper's data storage and event-driven features. The event highlighted the potential of Harper’s open-source platform and set the stage for future innovations, reflecting the dedication of developers and engineers who have contributed to Harper’s evolution over the years.

Ivan Judson, a solution architect at Harper, offers insights into managing local instances with Fabric and deploying basic applications, emphasizing the flexibility of Harper's platform. The session covers the installation of Harper using Node Version Manager and provides guidance on managing local development environments. Additionally, Judson demonstrates deploying a simple static website, showcasing the practical applications of Harper's tools.

A major North American fashion retailer, experiencing slow website performance that affected user experience, SEO, and revenue, partnered with Harper to modernize its edge architecture and optimize caching, resulting in significant speed and caching improvements. The retailer initially faced challenges with long load times and poor Core Web Vitals, particularly on product detail pages, due to inefficient caching practices that caused high latency and infrastructure costs. Harper facilitated a rapid transition to a new edge platform, extending cache durations and implementing full edge replication to enhance performance globally while reducing origin load. Additionally, Harper deployed HTTP 103 Early Hints to improve visual rendering times by allowing browsers to preload critical assets. This collaboration led to a 70% reduction in load times, improved Core Web Vitals, and increased conversion rates, demonstrating how strategic optimization of caching and content delivery can enhance user experience and drive business growth for e-commerce retailers.

Austin and Marianne discuss a pre-rendered demo built on the Harper platform, focusing on its architecture, rendering process, and development challenges. The demo enhances SEO by serving pre-rendered content efficiently, utilizing Puppeteer for rendering and orchestrating jobs across nodes. They emphasize the importance of performance optimization and managing edge cases, revealing that the system can handle 2,000 pages per minute. The architecture features a delivery cluster and render service, ensuring atomicity in the database and preventing race conditions in job processing. The orchestrator node is key in job management, and future improvements will incorporate community feedback. The demo is already in use with several customers, although public display is pending permission.