We're releasing a visual agent & workflow builder Fully open source Built on http://useworkflow.dev Outputs "𝚞𝚜𝚎 𝚠𝚘𝚛𝚔𝚏𝚕𝚘𝚠" code Supports AI "text to workflow" Powered by @aisdk & AI Elements Sample integrations (@resend, @linear, @slack) Clone & ship your own product, or embed AI workflow building capabilities into existing ones. Demo: http://workflow-builder.dev Deploy: http://vercel.com/templates/ai/workflow-builder…

What’s your preferred agentic engineering interface?

Learn to ship. Shipping is a skill distinct from coding. Shipping is designing, coding, QAing, story-telling, teaching, marketing, selling, pivoting, iterating… It used to be that coding dominated in importance because of coding ability scarcity. AI will push you to go further.

Can ChatGPT run Doom? Yes. ChatGPT Apps are very powerful. I cloned our Next.js ChatGPT template, registered a 𝚙𝚕𝚊𝚢_𝚍𝚘𝚘𝚖 MCP tool and deployed to @vercel . Once the tool is called, ChatGPT embeds the full @nextjs application. Server and client rendering just works, and it's 100% interactive. h/t @andrewqu_ @allenzhou101 for the starter kit: http:// vercel.com/templates/ai/c hatgpt-app-with-next-js …

Opus is on a different level. It's unreasonably good at @nextjs and the best model we've tried on @v0 to date. This is a one-shot generation. For a limited time you can try it on http://v0.app at no extra cost

We should not try to make AI agents into something more complex than they are. Technically speaking, an agent is a workflow. Many agents are crons, of which workflows are a superset. Because agents are workflows, we’ve seen a renaissance of workflow builders: Zapier, n8n, etc Whether you build an agent visually or with code, underpinning the workflow there must be an engine and infrastructure that makes it reliable. A workflow is a different kind of software than a web server. It’s not request-response. It consists of many steps, can fail a lot and is expected to recover, it can run for a very long time. This is why we’re so excited about @WorkflowDevKit . An agent is just “𝚞𝚜𝚎 𝚠𝚘𝚛𝚔𝚏𝚕𝚘𝚠”. Pair it with @aisdk (where each tool call is an “𝚞𝚜𝚎 𝚜𝚝𝚎𝚙”), AI Gateway for token reliability, and Fluid for efficient compute… and you have all you need to build the future of software.

The long way is actually the shortcut

The reason “vibe coding” continues to grow and be successful is that the alternative to vibe coding is not “elite engineering”. It’s: the project wasn’t born, the idea didn’t get communicated, the app didn’t ship. Elite engineering is very scarce and will continue to be in extremely high demand. (We’re hiring elite engineers!) The gap between what top engineers and agents can do still exists. Not just that.. when those people use AI, they also gain superpowers.

The world runs on TypeScript & JavaScript. Our bet is that AI engineering will follow suit. The growth in @aisdk downloads and adoption has been astonishing. When we wrote the Ship AI keynote it was at 3.4M weekly downloads. A couple weeks later, it’s now at 4.1M https:// vercel.fyi/ai-npm

Vibe coding vs Vibe engineering: Using AI to operate on the 𝘱𝘳𝘰𝘥𝘶𝘤𝘵 Using AI to operate on the 𝘤𝘰𝘥𝘦 The former is vibing the later is engineering. Obviously you might be thinking: “the code is the means to operate on the product! These are the same!” I believe on a long enough lifetime, for the vast majority of software, they will be. Products like @v0 and @lovable_dev exist in the former category. At @vercel however our heart is in engineering the best possible products, with the highest possibly quality, which means also having clean ejection and interoperability points with formal engineering tools. I believe there’ll be ample success and large markets for both vibing and engineering. It’s very likely vibing will become more *opinionated* over time, to make the reliability, security and quality bar *everyone* wants. As an example, just yesterday I watched a demo of a highly opinionated full-vibing platform being built on Vercel that delivers better outcomes than engineers, due to the constraints and opinions it imposes. AI in the service of slop is a temporary, “AI interest rates phenomenon” (AIRP), a period in time where people paid for the fascination of using AI even though the outcome looks like deformed Will Smith in software form. Our focus at @vercel continues to be AI in the service of quality, not just productivity alone.