April 2026 Summaries

At Vercel's Demo Day, 39 teams from the 2026 cohort showcased their AI innovations developed during a six-week intensive program at the company's San Francisco headquarters. The program, which included technical workshops and fireside chats with industry leaders, aimed to accelerate the development of AI applications in fields like finance, security, healthcare, and robotics. Participants received over $200K in infrastructure credits and support from partners like AWS and OpenAI. Demo Day culminated with pitches to investors and AI leaders, where judges evaluated teams based on criteria like technology fit and product quality. The top three teams were awarded additional credits and prizes, with Rex, an AI solution for enterprise finance operations, securing first place and receiving an investment from Vercel Ventures. The program's alumni have collectively raised over $100M in venture funding, highlighting the accelerator's role in fostering rapid innovation and collaboration within the AI startup ecosystem.

Zo Computer, a personal AI cloud platform co-founded by Rob Cheung, aims to democratize cloud computing by providing users with their own AI-enabled cloud environments. This innovative approach allows individuals to manage various tasks such as business operations, research, and health data tracking through a seamless, conversational interface. Transitioning to Vercel's AI SDK and AI Gateway significantly improved Zo's infrastructure by reducing code complexity and enhancing reliability, leading to a dramatic reduction in retry rates and latency issues. This shift enabled the small team to focus on scaling their user base, targeting the onboarding of one million users by 2026, while ensuring a responsive and efficient user experience. The partnership with Vercel has allowed Zo to handle increased traffic with ease, maintaining high performance and user satisfaction.

Vercel Workflows is a powerful tool designed to facilitate the creation of long-running, reliable, and observable agents and backends by integrating the orchestration of durable workflows directly into application code. Since its beta launch in October 2025, Workflows has handled over 100 million runs and 500 million steps, demonstrating its scalability and efficiency across a wide range of applications, from AI agents to complex backend processes. It operates without a separate orchestration service, running entirely within the application code, which simplifies deployment and reduces overhead. Workflows is compatible with both TypeScript and Python, offering a seamless experience for developers by managing queues, retries, state, and streaming through a small API surface. Additionally, it provides built-in security features, such as automatic data encryption, and supports durable streams for resilience against interruptions. Companies like Mux, Durable, and Flora have successfully implemented Vercel Workflows to streamline their AI and media processing services, emphasizing its role in reducing infrastructure complexity and enhancing product development. With the general availability of Workflows and the upcoming release of Workflows 5, Vercel aims to further improve performance and integration, making it an attractive option for developers seeking to build scalable and secure applications.

GitBook, an AI-native documentation platform, hosts 30,000 documentation sites on Vercel, serving 120 million monthly page views, with 41% of traffic from AI crawlers. The platform's integration with Vercel has addressed the challenge of keeping documentation fast and current by utilizing tag-based cache invalidation, which ensures that only affected content is revalidated during updates. This approach has enabled GitBook to process 40,000 cache invalidations daily within 300 milliseconds, supporting its commitment to providing fast and reliable documentation akin to production code. The rise in AI-driven traffic, which surged fivefold in 2025, has necessitated robust caching strategies to handle unpredictable traffic patterns while maintaining consistent content and predictable infrastructure costs. As GitBook continues to scale, it aims to enhance its caching infrastructure to accommodate adaptive documentation features and the increasing volume of requests, ensuring low latency and high predictability in delivering documentation.

Cursor undertook a significant transformation by launching a new brand and redesigning their marketing site to unify four web surfaces under a single domain, cursor.com, while maintaining essential signup flows. This initiative, executed using Microfrontends, allowed for a seamless transition without downtime and expanded their localization from four to eleven languages, resulting in a 5% increase in product-led growth signups. The company adopted an agent-first workflow, moving away from traditional CMS to a system where content updates are made directly in code, leveraging cloud agents in Slack for requests and GitHub for deployment previews. This approach not only streamlined operations but also democratized participation across the team, allowing for rapid, iterative improvements supported by data-driven experimentation using tools like Statsig and Vercel Flags SDK. This strategic overhaul has positioned Cursor for scalable growth, allowing them to experiment with content and user flows confidently, enhancing both their product-led and sales-driven initiatives.

The evolution of infrastructure is moving towards a new era driven by large language models (LLMs) and coding agents, which are increasingly responsible for software development, testing, and deployment. This shift is evidenced by the rapid increase in agent-initiated deployments, with Vercel witnessing a significant rise in weekly deployments driven by agents. As software becomes more agentic, requiring infrastructure that can autonomously manage, adapt, and respond to changes, companies like Vercel are developing platforms that unify AI capabilities, providing a seamless environment for coding agents to operate. This includes features such as programmatic deployments, agent-oriented serverless workloads, and enhanced observability, aiming to create a system that not only supports but also acts as an agent itself. The infrastructure is designed to handle the complex needs of AI workloads, ensuring efficiency and resilience while reducing human intervention, ultimately positioning itself as a proactive participant in software management.

AI Gateway simplifies data policy management for teams using multiple AI models by providing automated enforcement of Zero Data Retention (ZDR) agreements, thereby reducing the manual, error-prone processes typically involved in managing fragmented data policies across different providers. This service negotiates ZDR agreements with providers such as OpenAI, Anthropic, and Google, ensuring that data requirements are automatically met without the need for code changes, either on a team-wide or per-request basis. AI Gateway also offers explicit controls to disallow prompt training, protecting proprietary and sensitive data from being used in model training, and provides a comprehensive audit trail of data policy enforcement. This infrastructure-centric approach allows teams to focus on development rather than compliance, ensuring that data protection measures are integrated seamlessly into their workflows.

Vercel recently improved the performance of its filesystem snapshots in Vercel Sandbox by significantly reducing snapshot restore times from over 40 seconds to under one second. Initially focused on reliability, the team faced performance challenges when snapshot restore times were lengthy, prompting them to implement parallelization and local caching. By leveraging the AWS Go SDK's transfermanager API and performing parallel S3 range requests, they accelerated download times by 2-5 times. Additionally, they enhanced decompression speed by using multiple goroutines, further reducing restore times by 2-4 times. The team also introduced a local disk cache using LRU eviction, achieving a 95% cache hit rate, which allowed most snapshot restores to bypass both download and decompression entirely. These optimizations have not only improved snapshot restore performance but also support Automatic Persistence, ensuring an instantaneous experience when stopping and resuming sandboxes.