How do Vercel Workflows allow for long-running processes?

Vercel Workflows enable serverless functions to be indefinitely long-running by allowing code to pause and resume. This means you can wait for multiple days for tasks like human approval via webhooks, without incurring extra costs.

What is Vercel's open-source business model strategy?

Vercel employs a model where their software library is truly open source and runs everywhere, making it popular. They then capture a piece of the overall growing market, aiming to grow the pie for everyone involved.

Why has Vercel's AI SDK been successful?

The AI SDK's success stems from Vercel's humble approach, staying low-level to remain flexible as AI trends emerge. They avoided thick abstractions, allowing it to adapt from chatbots to agents without a rewrite, and it forms the often unseen basis for other frameworks.

What is the difference between Vercel's internal agents and their productized agents?

Internal agents are used to improve Vercel's own operations, like in their DevOps or lead qualification processes. Productized agents, like the Vercel Agent, are 'agent as a service' offerings designed for customers to build applications on Vercel, leveraging Vercel's ecosystem data.

What types of problems are current AI agents best suited for?

AI agents are effective for problems that are tedious and repetitive, requiring some human judgment but are not easily automated otherwise. They excel in tasks like processing sales requests, abuse analysis, and data analysis, often identifying 'sweet spots' that have high business impact.

How is Vercel supporting Python development?

Vercel has invested significantly in Python support, offering zero-config deployment for popular frameworks like Flask and FastAPI. They also provide a Python SDK for their API and are expanding their 'fluid compute' program to better support Python workloads, especially for AI models.

What leadership principles guide Vercel's CTO in the AI era?

A key principle is ensuring that transformations, like embracing AI, feel native to the company's existing strengths. Malte Ubl also emphasizes developing processes that assume developer and AI incompetence to build inherently more secure applications, moving beyond simple trust models.

⚡️ Ship AI recap: Agents, Workflows, and Python — w/ Vercel CTO Malte Ubl

Latent Space Podcast

Science & Technology4 min read43 min video

Nov 1, 2025|2,257 views|51|4

Save to Pod

Key Moments

TL;DR

Vercel CTO Malte Ubl discusses company's AI strategy, focusing on workflows, SDKs, and agents. Emphasizes dogfooding and open-source contributions.

Key Insights

Vercel's AI strategy prioritizes practical application through "dogfooding"—using their own tools internally before releasing them.

The new workflow development kit simplifies complex processes, making workflows a first-class, everyday development task.

Vercel's AI SDK remains low-level and flexible, adaptable to emerging AI trends rather than imposing rigid abstractions.

Vercel supports open-source with a model where broad adoption benefits the ecosystem and Vercel's growth.

The Vercel agent ecosystem aims to integrate AI into development workflows, with products like a DevOps agent for anomaly detection and a lead qualification agent.

The company is investing in Python support on Vercel, offering zero-config deployment for popular Python frameworks.

VERCEL'S COMMITMENT TO AI ENGINEERING AND PRACTICAL APPLICATION

Vercel, through its CTO Malte Ubl, positions itself as a strong advocate for the AI engineering movement, emphasizing practical application over hype. The company's philosophy centers on "dogfooding"—rigorously using their own tools internally to ensure they are battle-tested and provide real value before being offered to customers. This approach grounds their product development in real-world use cases, allowing them to extract and refine useful abstractions. This internal validation is key to their strategy for building production-ready agents and efficient internal operations.

REVOLUTIONIZING WORKFLOWS FOR MODERN DEVELOPMENT

A significant announcement from Vercel's Ship AI event was the new workflow development kit, designed to make workflow creation an integrated and intuitive part of daily development. Workflows, often an overlooked concept in traditional computer science education, are presented as crucial for managing complex processes. Vercel's kit enables workflows to be durable, zoomable, and streamable, allowing code to pause and resume indefinitely without cost, a paradigm shift from traditional serverless functions. This significantly enhances reliability, especially for long-running orchestration tasks.

THE AI SDK'S FLEXIBILITY AND OPEN-SOURCE PHILOSOPHY

Vercel's AI SDK, a successful open-source project, is characterized by its deliberate low-level design. This approach ensures flexibility in a rapidly evolving AI app space where specific use cases are still emerging. By avoiding thick, premature abstractions, the SDK can adapt to new trends, such as the shift from chatbots to agents, without requiring major rewrites. Vercel champions an open-source business model where broad adoption of their libraries makes them popular, allowing Vercel to capture a proportional share of growth while benefiting the entire ecosystem.

INTEGRATING AGENTS INTO THE DEVELOPMENT LIFECYCLE

Vercel is developing an ecosystem of AI agents designed to assist developers in building applications on the Vercel platform. These agents leverage Vercel's unique access to runtime data, error logs, and deployment information to provide integrated solutions. Examples include a code review agent and a broadly applicable DevOps agent that investigates production anomalies by querying observability data and logs. The goal is to offload tedious tasks, allowing agents to handle investigations that might otherwise consume significant developer time, particularly in nuanced situations like anomaly detection where they can operate without human disturbance.

THE VERSEL AGENT ECOSYSTEM AND CUSTOM AGENT DEVELOPMENT

Beyond productized agents, Vercel is fostering custom agent development through initiatives like the 'Agent on Every Desk' program. This program offers support, including forward-deployed engineers, to help companies, especially larger ones, overcome the initial hurdles of building their first agents. The focus is on identifying high-impact, yet tedious, tasks that current agents can reliably handle. Vercel has open-sourced internal agents for lead qualification, abuse analysis, and data analysis, providing developers with practical starting points for creating their own tailored solutions.

EXPANDING INFRASTRUCTURE SUPPORT FOR PYTHON AND OTHER LANGUAGES

Vercel is actively investing in robust support for Python, recognizing its importance in the AI ecosystem. They now offer zero-config support for popular Python frameworks like Flask and FastAPI, allowing seamless deployment. This expansion extends to a Python SDK for their API. While Vercel historically focused on JavaScript, their infrastructure, built on virtual machines, can accommodate various languages. The key is ensuring a high-quality developer experience (DX) that feels native to each ecosystem, a process that requires substantial investment in engineering and infrastructure.

THE EVOLVING CTO ROLE AND SECURE APP DEVELOPMENT

Malte Ubl reflects on the transformation of the CTO role, particularly with the advent of AI. Vercel's pivot to AI felt native, leveraging their expertise in frameworks. A critical focus for Vercel is building applications with robust security models, assuming developers and AI might not always be competent. They are developing systems to ensure security is independent of developer skill, extracting sensitive operations from applications to prevent potential misconfigurations. This forward-thinking approach aims to enable secure app development, even with less experienced teams.

Mentioned in This Episode

●Products

●Software & Apps

●Companies

●Organizations

●Concepts

●People Referenced

Common Questions

Vercel's message is to be the biggest fan of the AI engineering movement by focusing on concrete actions and building practical tools, rather than just hype. They aim to make building with AI easier and discover the right abstractions as the field emerges.

Topics

Serverless Functions Anomaly Detection CTO App Security

Mentioned in this video

organizationF.NG

The organization that provided the remote studio space for the recording.

softwareM-Wiz

A system Malte Ubl worked on at Google, which is no longer publicly accessible.

conceptShip AI

Vercel's conference focused on AI engineering, recapped in this discussion.

productWorkflow Development Kit

A new product Vercel shipped, designed to make writing workflows first-class and idiomatic.

organizationC39

An organization Vercel is working with to integrate decorators into more places.

productVercel Data Analyst Agent

An internal Vercel agent designed to query data warehouses, open-sourced.

productAgent on Every Desk Program

A Vercel initiative offering 'forward-deployed engineering' to help companies build their first agents.

productDevOps Agent

A Vercel agent announced that ties into their anomaly detection system for investigating production issues.

productVercel Agent for Lead Qualification

An internal Vercel agent that processes incoming sales requests, open-sourced for others to use.

productVercel Agent for Abuse Analysis

An internal Vercel agent that performs pre-work for abuse reports, assisting human review.

softwareWork OS

Mentioned as an example of a system that has approached Vercel's goals for secure, agent-native infrastructure.

softwarev0

toolTemporal

toolSentry

Found this useful? Build your knowledge library

Get AI-powered summaries of any YouTube video, podcast, or article in seconds. Save them to your personal pods and access them anytime.

Try Summify free