December 2025 Summaries

Anticipatory customer experience (CX) represents a shift towards predictive, proactive engagement, enabled by real-time data infrastructure that allows companies to address customer needs before issues arise. This approach relies on continuous data streams and event-driven patterns to replace traditional reactive models, offering interventions before problems occur. Industries such as healthcare, retail, and financial services are leveraging real-time platforms to foresee and mitigate potential customer dissatisfaction by providing timely updates and targeted support, thus enhancing trust and reducing friction. Companies like Doxy.me exemplify this transition by integrating real-time features that anticipate customer confusion and streamline processes. The success of anticipatory CX hinges on reliable infrastructure, with systems like Ably's ensuring uninterrupted and adaptive service delivery, which is crucial for maintaining customer trust. As customer expectations evolve, businesses are increasingly investing in event-driven architectures that allow for scalable and flexible foresight, positioning them advantageously in what is termed the Age of Anticipation.

Ably, a company known for delivering reliable, low-latency real-time experiences at scale, has evolved to address the challenges of modern AI-driven interactions with its new Ably AI Transport Layer. This layer is designed to support "Gen 2" AI experiences, which involve complex multi-step workflows, asynchronous operations, and continuous feedback, moving beyond the traditional request-response pattern. Ably AI Transport focuses on providing a robust infrastructure between AI agents and user devices, allowing for seamless cross-device continuity, state recovery, and multi-agent coordination. It introduces features like shared state using CRDT-based collaborative objects, enabling sophisticated coordination and context-aware prioritization among AI agents. By maintaining a framework-agnostic approach, it allows developers to focus on AI logic and orchestration without being constrained by networking issues, offering integrations with platforms like OpenAI's agent framework and LangChain. Currently, Ably is refining these concepts based on real-world usage while advocating for a streamlined approach to building AI experiences without the need to reconstruct real-time infrastructure.

As AI becomes increasingly integrated into everyday applications, users now expect seamless interactions, where AI responses continue uninterrupted despite technical disruptions such as page reloads or network drops. Reliable and resumable token streaming is essential to meet these expectations, ensuring that AI conversations can pick up exactly where they left off without restarts or lost context. This requires a robust transport layer that maintains persistent streaming connections, buffers the AI's output server-side, and tracks session states across client restarts. However, implementing such a seamless experience poses challenges due to the limitations of traditional web protocols like HTTP, which are not designed for long-lived, continuous streams. The solution lies in infrastructure capable of handling reconnections and delivering ordered, exactly-once messages, transforming the web into a reliable conduit for AI data. By addressing these technical hurdles, platforms like Ably AI Transport aim to provide the necessary infrastructure for developers to deliver continuous, uninterrupted AI experiences across devices and sessions.

Cross-device AI synchronization enables conversations to maintain continuity across multiple devices, allowing users to start a chat on one device and seamlessly continue on another without losing context or having to reintroduce themselves. This technology builds trust and reduces friction by maintaining a persistent presence, making AI assistants more reliable and useful. However, achieving this seamless experience is technically challenging due to the stateless nature of HTTP, which requires sophisticated infrastructure to handle message synchronization, ordering, and session recovery across devices. Implementing a robust transport layer with features like persistent bi-directional messaging, identity-aware fan-out, and real-time streaming is essential for ensuring smooth AI interactions. Ably's platform provides these capabilities, enabling developers to focus on enhancing AI functionality rather than building the underlying communication protocols, thereby facilitating the development of synchronized AI experiences across devices.

Ably AI Transport is a new early-access platform designed to enhance the functionality and reliability of AI models by providing a seamless, continuous, and multi-device user experience. It addresses the limitations of current AI infrastructure, which often relies on short-lived HTTP requests that hinder the potential of advanced AI models. By using a global WebSocket platform, Ably AI Transport ensures the continuity and shared state of AI interactions, allowing for uninterrupted, real-time communication and collaboration across devices and sessions without requiring changes to the existing tech stack. This innovation supports the deployment of next-generation AI applications with improved user experiences, offering a stable and reliable foundation that adapts to evolving models and frameworks.

Faced with the discontinuation of HashiCorp's CDK for Terraform, the Infrastructure Team at Ably decided to create their own tool, tfgen, using Go, to replace CDKTF's functionality while maintaining their existing infrastructure processes. The team sought a solution that would not necessitate extensive migration or disrupt their current workflows, leading them to develop tfgen, which generates Terraform-compatible JSON without requiring strongly typed provider definitions. This approach allowed them to bypass the limitations of CDKTF and streamline the configuration process. By leveraging AI for the migration process, they managed to transition all their workspaces efficiently and have open-sourced tfgen to benefit the community, providing a flexible and provider-agnostic tool to generate Terraform configurations.