April 2026 Summaries

The text discusses the evolving landscape of production AI, emphasizing the necessity for a session layer in addition to streaming to enhance AI user experiences, especially as AI tools transition from internal to customer-facing applications. The AI Engineer Europe event highlighted a shift towards building sophisticated AI products that prioritize user experience, with innovations like multi-agent systems and interactive user interfaces. A key takeaway is the importance of quality and resilience in AI systems, which is challenged by current HTTP streaming architectures that falter in maintaining continuity and control across various devices and connection drops. The text proposes that teams should adopt durable sessions, a concept where a stateful layer between agent and client ensures persistent connectivity and control, allowing for seamless reconnection and interaction across multiple devices. This approach, illustrated by Ably's AI Transport, decouples token generation from delivery, thus maintaining user control and improving the reliability of AI products.

The text provides a brief overview of a blog post section, highlighting that it includes various articles related to security tests and updates on April 25, 2026, as well as a discussion on the development of the Durable Sessions stack on April 24, 2026. It also encourages readers to subscribe to the Ably newsletter for monthly insights on the real-time data economy, emphasizing the trust industry pioneers place in Ably for such information.

This brief text highlights a blog post update from April 25, 2026, mentioning Matthew O'Riordan as the author and listing recommended articles related to security testing, specifically SSRF and OOB tests. It also references a longer article titled "The Durable Sessions stack is forming," published on April 24, 2026, and invites readers to join the Ably newsletter for monthly insights on the realtime data economy, emphasizing its trust among industry pioneers.

Matthew O'Riordan's post, last updated on April 25, 2026, mentions several updates and tests related to security, specifically addressing SSRF (Server-Side Request Forgery). The blog also highlights an ongoing development in AI transport and mentions the formation of the Durable Sessions stack as of April 24, 2026. Additionally, it promotes the Ably newsletter, which provides monthly insights into the realtime data economy, trusted by thousands of industry pioneers.

Durable Sessions are emerging as a crucial infrastructure component in AI systems, providing resilience by ensuring continuity despite potential failures across various layers of operation. This concept, which includes durable memory, execution, and transports, is seen as essential for enabling long-lived AI systems to maintain functionality through crashes, retries, and disconnections. The idea has gained traction as companies like Ably and ElectricSQL explore its potential, each offering different approaches based on their expertise in real-time communication and database management, respectively. Durable Sessions encapsulate the persistence of user interactions with AI agents, ensuring seamless transitions across devices and maintaining state even when offline. This approach is supported by a broader ecosystem of companies contributing to the development and implementation of durable infrastructure, highlighting the importance of reliable and consistent AI experiences. The evolving conversation around Durable Sessions suggests a growing recognition of the need for robust session handling in AI applications, with ongoing contributions and discussions shaping its definition and application across the industry.

Ably has launched version 3 of its Python SDK, designed to enhance AI infrastructure by addressing challenges in production environments where Python dominates AI development. This release focuses on improving real-time data streaming with AI Transport support, ensuring token streaming with ordering guarantees, and enabling resumable sessions to maintain continuity despite network issues or device changes. Multi-device continuity is facilitated through channel-based sessions, allowing users to switch devices without losing conversation history. The SDK offers live steering, enabling users to redirect responses mid-stream, and supports real-time publishing over WebSocket connections to improve efficiency compared to REST calls. Additional features include presence tracking to manage agent activity based on user engagement and mutable messages that allow editing post-publication for more efficient history management. The update aims to establish Python on par with other real-time SDKs and plans further enhancements like LiveObjects support for shared state synchronization, with the Ably team seeking user feedback to guide future developments.

Matthew O'Riordan discusses the growing need for durable session infrastructure in AI applications, highlighting the common challenges faced by engineering teams when AI agents lose session continuity due to factors like users switching devices or experiencing connectivity issues. Despite the focus on AI intelligence and orchestration, the transport layer has become a bottleneck as AI agents increasingly mimic human communication. Companies like Intercom and HubSpot have already leveraged existing messaging infrastructures such as Ably's Pub/Sub to enhance AI-human interaction, while others like Vercel and EMQX are developing solutions to address these infrastructure gaps. The concept of "Durable Sessions," which ensures session persistence across devices and disruptions, is gaining traction, mirroring the evolution of Durable Execution in backend workflows. O'Riordan emphasizes the importance of industry collaboration to establish durable sessions as a recognized infrastructure layer and invites stakeholders to engage in discussions to advance this emerging category.

Amber Dawson's article addresses the challenges and solutions surrounding multi-device AI session continuity, specifically highlighting the difficulties encountered when users switch devices during AI-driven interactions. The core issue stems from the stateless nature of HTTP streaming, which does not support cross-device session continuity, leading to broken conversations, duplicate efforts, and state conflicts. To overcome these challenges, the article suggests a shift to a channel-based architecture where session state is maintained in a channel rather than tied to individual connections. This enables seamless reconnection, message history replay, and consistent updates across devices, ensuring that users can pick up conversations where they left off. The article also emphasizes the complexity of mobile environments, which require additional strategies like push notifications and intelligent reconnection to address network instability and battery constraints. Ultimately, it argues that for Gen-2 AI applications, which involve longer sessions and multi-device interactions, a channel-based architecture is essential to meet user expectations and maintain session integrity.