Your @OpenAI agent's answer now survives a refresh or a dropped connection, because the reply lives on a session instead of on the connection. Any device can pick the same thread back up. Demo: ably.com/blog/ai-transport-o… #AIAgents

A user refreshes halfway through an AI response. What comes back? Store tokens as separate events, and every client has to rebuild the answer. We compared four ways to store streamed AI responses, and how Ably’s message appends solve it: ably.com/blog/store-streamed…

WebSockets let users interrupt or approve an AI agent mid-stream. They don’t handle session recovery, fan-out or delivery guarantees on their own. When does running that layer yourself make sense? Here's a breakdown of the technical and cost trade-offs: ably.com/blog/websockets-bea…

Guaranteeing WebSocket order in production isn't a one-time fix. TCP's guarantee ends at one connection. So past that, sticky routing, a shared coordination layer, and replay sequencing are system requirements. But should you manage the infra? ably.com/blog/websocket-orde…

Fin's Colin Kennedy, on deciding whether to keep maintaining your own WebSocket infra for AI streaming: "Is maintaining this system helping you build a better product?" The framework to work through that decision: ably.com/blog/websockets-ai-…

Building LLM streaming? You've probably got session IDs, a Redis buffer, and dedup logic. That's five systems you maintain forever, and there's a sneakier failure too: streams that look fine while nothing confirms delivery. The fix: ably.com/blog/resume-tokens-… #AIEngineering

Your AI agent goes quiet mid-task. Thinking, stalled, or dead? From the client, you can't tell. That's because most AI infrastructure has no signal for it, as Vercel's own maintainers have said. Ably AI Transport gives you one. ably.com/blog/ai-agent-model…

Three agents check flights, hotels, visa rules in parallel. Your UI: one status line, "Working on it..." That's structural. Change a hotel mid-search, and the UI won't show it until the orchestrator reports back. See what real collaboration needs. ably.com/blog/multi-agent-ai…

Presence and shared state used to mean building your own layer for it. AI Transport v0.3.0 builds presence in: one call answers whether the agent is online and who else is watching. LiveObjects, opt-in, syncs shared state too. See what shipped, with code. ably.com/blog/introducing-ai…

Your AI chat demo works ✅ Then you ship it - and a refresh mid-response restarts the conversation 🔄 HTTP streaming hits limits past simple prompt-response: - Lost context on refresh - No agent visibility - Can't interrupt a stream See how a session layer fixes this - and when you need to introduce one. ably.com/blog/ai-stack-sessi…

Model providers handle inference. Agent frameworks handle orchestration. Temporal makes the backend crash-proof. Every layer has a name. Except for the one between the agent and the user's device. Discover what this looks like 👇 ably.com/blog/stateful-agent…

Workflow Streams lets a client watch output. It can't send anything back. Fine, until someone needs to step in mid-task. There's no way in: redirecting means starting over and losing progress. We break down where Workflow Streams ends and what needs to sit alongside it. ably.com/blog/temporal-workf…

A third person joins an AI conversation an hour in. In most products, they're starting from zero. A durable session catches them up automatically: full history, no special load-history call, same mechanism as a page refresh or device switch. We show exactly how this works in a real demo. Worth seeing end to end 👇 ably.com/blog/durable-sessio…

A vanished AI conversation on reload is a support ticket, a lost upsell, or a compliance question. useChat doesn't persist history by default. A database write fixes reload, not a second device joining live. We cover when a database write is enough, and when it isn't. ably.com/blog/ai-chat-stream…

Most teams don't choose their AI chat transport. It just comes with useChat. It works fine, until a real user does something the demo never covered: a tab switch, a second device. We mapped exactly where the gaps show up. Worth checking before you find out the hard way. ably.com/blog/custom-transpo…

Your Java backend tracks a leaderboard. Someone resets it. Your code doesn't know, and now needs manual rebinding. LiveObjects for Java fixes this: point at a location, not a specific object. It resolves to whatever's there now. Full examples in Java and Kotlin. ably.com/blog/liveobjects-ja…