Observed Signal · Jul 3, 2026 · Interview · Source: AINews swyx · Impact: 2/5 · Sentiment: Neutral
Vercel's Andrew Qu: Agents as a New Software Class
Andrew Qu, Chief of Software at Vercel, explains in an interview why AI agents represent a new form of software, how Vercel built internal agents and turned the lessons into the eve agent framework, and why Vercel is making its platform agent-friendly. Qu describes technical primitives (skills, filesystem agents, resumability, sandboxes, subagents), use cases for business automation inside Vercel (contract redlining, marketing retrospectives, query writing), and the idea of serving machine-readable content (Markdown) to agents. He also highlights ongoing challenges—secure code execution, long-running jobs, and collaborative or "multiplayer" agent development—and says Vercel intends to integrate observability and partner integrations while keeping developer onboarding simple.
Platform-level developer tooling and agent frameworks influence how web properties and developer workflows evolve, with implications for machine-readable web traffic and automation, but this is a developer/platform update rather than an industry-shifting policy or major platform technical release.
Track Vercel Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Andrew Qu is Chief of Software at Vercel and leads internal engineering, product experimentation, and emerging-technology initiatives.
- Qu led development of eve, Vercel’s framework for building agents, which grew from internal tooling and reusable primitives.
- Vercel uses agents internally for repetitive business tasks including initial contract redlining, marketing retrospectives, contact identification, and data-store query generation.
- Vercel detects when an agent makes a website request and can serve Markdown (a machine-readable format) to agents instead of visual HTML.
- Qu built supporting tools such as skills.sh and an MCP library while iterating on agent patterns like filesystem agents, skills, compaction, and subagents.
Connected Companies & Entities
1 Entity mapped“Andrew Qu is Chief of Software at Vercel, where he works with the CTO across internal engineering, product experimentation and emerging tech...”
Ontology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Vercel CEO: Separate Models from Agents, Push Open Protocols
Vercel CEO Guillermo Rauch discussed the company’s role in evolving AI agents and the need to separate models from agents to preserve modular software engineering. Vercel reports about 6 million daily deployments (roughly half from coding agents) and more than 1 trillion tokens flowing through its AI gateway each day. Rauch described internal tools — a natural-language instructions framework called Eve and a Vercel Sandbox for applying data and policy controls — designed to make agents safe and auditable in production. He said customers are moving from single-lab lock‑in toward plug-and-play stacks (model, harness, data platform, sandbox, gateway) using providers like OpenAI, Anthropic and a range of open models. Rauch warned platform advances from major labs can compete with infrastructure providers and reiterated Vercel’s aim to enable open protocols and act as foundational infrastructure for agent-driven applications.
Vercel Uses AI Agents to Automate Tasks Without Layoffs
In a podcast interview, Julian Eckerle, Head of Central Europe at Vercel, discussed the company's internal adoption of AI agents. Vercel began integrating AI about two and a half years ago, encouraging all employees to build their own agents. This bottom-up approach led to numerous use cases across content, customer support, and lead processing. An autonomous sales development representative (SDR) now manages the entire inbound sales funnel, a task that previously required over ten people, with annual operating costs of $5,000. A customer support agent now automatically resolves 91% of tickets. Despite these automations, Vercel reports no layoffs; instead, employees have been moved to higher-value tasks like outbound sales, and the sales organization has more than doubled. The company also implemented an AI gateway to manage token costs and controls spending by using open-source models where appropriate. Eckerle emphasized that AI should create value for employees, not just more work, and cautioned against a blanket 'just use AI' approach without proper support.
Guillermo Rauch on Agent-Led Growth and the Software Factory
Guillermo Rauch, founder of Vercel and creator of Socket.IO and Next.js, spoke with a16z’s Gabriel Vasquez about the rise of AI agents and what he calls an “inexhaustible IQ machine.” Rauch described how companies can build a “software factory” that continuously learns and ships improvements, how Vercel consolidated internal automation into a single company-wide agent called V, and how agents already drive product discovery and purchases (what he terms “agent-led growth”). He emphasized that defensibility will come from the systems and verification loops built around models rather than from the models themselves.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.