Engineering Insights

Building Slack agents can be easy

Slack is already where teams work. It provides a natural interface for agents, with messages, threads, buttons, and events, so you don't need to invent a new UI or onboarding flow. Getting from "I want a Slack agent" to a running deployment, though, means coordinating across a lot of systems: Each piece has its own docs, and they all need to work together. Coding agents like Claude Code, OpenCode, Cursor, and GitHub Copilot are well suited for exactly this kind of coordination because they can r

Scaling redirects to infinity on Vercel

Redirects are trivial at a small scale, but at millions, latency and cost become real systems problems. Previously on Vercel, redirects were handled by routing rules and middleware. Routing rules support up to 2,000 complex redirects with wildcards, and they function as an ordered list evaluated in sequence. Each rule may involve regex matching, meaning a single request could trigger many expensive evaluations. This is acceptable for a few thousand routing rules, but as counts grow, per-request

Vercel Sandbox now accepts environment variables at creation

The Vercel Sandbox and now support setting environment variables at sandbox creation that are automatically available to every command.SDKCLI When running multi-step processes in a like installing dependencies, building a project or starting a dev server, each step often needs the same environment variables. Now, these are available with every call.Vercel SandboxrunCommand Environment variables passed to are inherited by all commands automatically. Per-command env in can still override ind

Vercel Workflow is now twice as fast

Server-side performance for Vercel Workflow, the fully managed platform built on top of the open-source Workflow Development Kit (WDK), is now twice as fast, delivering a 54% median improvement across the board. Over the last two weeks, the median API response time has been reduced from 37ms to 17ms, with queue latency, Time to First Byte (TTFB), and per-step overhead all reduced. Workflows that coordinate multiple steps benefit the most, as lower overhead compounds across each step in a run. To

GPT 5.3 Chat is now on AI Gateway

GPT-5.3 Chat (GPT 5.3 Instant) is now available on AI Gateway. This update focuses on tone, relevance, and conversational flow for more accurate answers, better-contextualized web results, and fewer unnecessary refusals and caveats. It also reduces hallucination rates and produces smoother and more direct responses. To use this model, set model to in the AI SDK.openai/gpt-5.3-chat AI Gateway provides a unified API for calling models, tracking usage and cost, and configuring retries, failover, a

Gemini 3.1 Flash Lite is now on AI Gateway

Gemini 3.1 Flash Lite from Google is now available on AI Gateway. This model outperforms 2.5 Flash Lite on overall quality, with notable improvements in translation, data extraction, and code completion. Gemini 3.1 Flash Lite is best suited for high-volume agentic tasks, data extraction, and applications where budget and latency are the primary evaluation constraints. To use this model, set to in the AI SDK. This model supports four thinking levels, , , , and .modelgoogle/gemini-3.1-flash-lite

Workflows, Workers - Workflows step limit increased to 25,000 steps per instance

Each Workflow on Workers Paid now supports 10,000 steps by default, configurable up to 25,000 steps in your wrangler.jsonc file: { "workflows": [ { "name": "my-workflow", "binding": "MY_WORKFLOW", "class_name": "MyWorkflow", "limits": { "steps": 25000 } } ] } Previously, each instance was limited to 1,024 steps. Now, Workflows can support more complex, long-running executions without the additional complexity of recursive or chi

GPT-5.3 Instant System Card