Benjamin Guo is an engineer and co-founder of Zo Computer, a personal cloud environment that brings AI agents, files, automations, and hosted services together. His work has included payment software, infrastructure for multi-step AI workloads, AI-assisted software development, and tools that let people build their own software. A recurring concern in his work is how much control people retain over their computing environments—and over the knowledge their agents accumulate.
From payments to AI infrastructure
Guo started his career on the early Venmo team in 2013, where he met his future co-founder, Rob Cheung. He later joined Stripe in 2015 and worked there for eight years. At Stripe, he worked on payment tools for developers. His 2017 iOS SDK update described reducing the backend endpoints required for Stripe’s prebuilt payment interfaces to one, alongside broader payment-method support, card scanning, address autofill, and customer-data prefetching. Guo has also been identified as the founding engineer of Stripe Terminal, Stripe’s in-person payments product.
After leaving Stripe, Guo and Cheung co-founded Substrate, an inference API for multi-step AI workloads. Its central abstraction was a computation graph: developers connected operations such as text generation, image generation, storage, and sandboxed code execution, allowing the backend to optimize the complete workload. Substrate’s architecture could colocate related computation, rewrite graphs, and run independent operations in parallel. Lightspeed led its $8 million seed round.
Zo Computer: a personal cloud for building software
Guo and Cheung subsequently shifted from their original agentic-inference product toward personal cloud computing. They reasoned that language models could make software creation available to people who had never considered themselves developers. Zo Computer had a major launch in late November 2025 after operating in public beta. In his account of the launch, Guo emphasized explaining what an intelligent cloud computer could do for someone like his mother rather than leading with its infrastructure. He also described recording the launch video’s background music in Ableton.
Zo puts AI, storage, connected tools, automations, and hosting in a persistent environment. Users can build websites, run scheduled tasks, work with files, or host services in the same place. In “Everyone Gets A Software Company”, Guo demonstrated a workspace with a filesystem, model chat, integrations, automations, and a browser. For technical users, he described a Linux virtual machine with root access, SSH, API and MCP access, and hosting for arbitrary HTTP or TCP services. His aim is to make that infrastructure useful without requiring everyone to learn deployment or server administration.
The customer stories in the talk illustrate what this model can change. Guo described a private chef and life coach using Zo for websites, invoices, bookkeeping, scheduling, and notes. He also described a freediving instructor connecting her retreat website with customer inquiries and business records. When someone expressed interest in a retreat, Zo texted her the prospect’s number; she could call immediately, request a payment link by text, and close the booking. Her database, accounting, notes, and website files remained in the same environment. These are Guo’s reported examples rather than independently verified customer outcomes, but they show his emphasis on reshaping operational workflows rather than merely generating standalone websites.
Ownership of software and agent intelligence
Guo’s criticism of subscription software centers on fragmentation and dependence. When booking, accounting, websites, and notes live in separate services, users must connect those systems while remaining subject to each provider’s pricing, product decisions, and data controls. He calls this structure “technofeudalism.” Personal cloud ownership is his proposed alternative: a place where people can keep their files, run agents, and operate services they have built. He connects that ambition to the early internet’s personal websites and individually shaped tools.
Messaging is part of this approach. Guo demonstrated texting his persistent cloud computer to act when an inquiry, purchase, idea, or task arose while the underlying environment continued running between interactions. In his own workspace, he showed a custom booking tool that allowed his agent to review requests before someone could book time with him.
His ownership argument extends to personal cloud agents. As agents acquire context or improve through use, he asks who benefits from that accumulated intelligence. In the talk, he advocated allowing individuals and companies to publish and own agents whose use improves their own systems. He described Zo’s agent-publishing system as a beta and did not specify whether improvement would occur through memory, revised instructions, model training, or another mechanism.
AI-assisted engineering
In “Velocity Coding, Not Vibe Coding”, Guo described AI-assisted software engineering as “leveraged thinking.” His workflow begins with manually locating relevant code and building familiarity with the codebase. He then uses a capable model to generate and repeatedly refine a detailed plan divided into implementation phases.
Once the plan is settled, Guo delegates implementation to a faster model. He generates changes incrementally, reviews each diff, runs tests and type checks, and commits each phase before proceeding. He compares working with coding agents to managing an engineering team: the person supplies context, decides which problems require human attention, delegates bounded tasks, and verifies the results. He also describes automated checks, conventions, and ratchets as a codebase’s “immune system,” particularly important when agents increase the speed and volume of changes.
Benjamin Guo presents Zo as a personal cloud: one AI-equipped server where an individual can keep files, run automations, publish websites and APIs, and replace parts of a fragmented SaaS stack. The larger question is who owns the cloud agents—and who benefits as those agents improve.
A personal cloud combines files, AI, hosted services, and local-device access around one individually controlled server rather than scattering the workflow across unrelated SaaS products.
The retreat workflow shows the concrete advantage of consolidation: a website inquiry triggers a text, the owner calls immediately, Zo creates a payment link, and the business records remain in the same environment.
Zo spans a plain-language workspace and a developer-accessible Linux server, with storage, automations, integrations, browser use, arbitrary service hosting, SSH, API, and MCP access.
Direct publishing removes deployment friction for nontechnical users, but the talk leaves safety, testing, rollback, and operational reliability largely unspecified.
The strategic question for cloud agents is where accumulated intelligence goes: toward the provider renting the agent, or toward the individual or company that owns and publishes it.
