Impressions from visiting OpenAI, Anthropic, & Cursor

2026-06-30 · Source: The Pragmatic Engineer · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Cloud Computing & IT Infrastructure · Depth: Advanced, medium

Summary

Recent visits to OpenAI, Anthropic, and Cursor highlight a significant industry shift towards cloud-based AI agents for long-running tasks, moving away from local machine execution. This "new paradigm" is evident in Anthropic's Claude Managed Agents, Peter Steinberger's Crabbox for OpenClaw, OpenAI's acquisition of Ona (Gitpod) for cloud development environments, and Cursor's Cloud Agents, which power its new iOS app. This transition is driven by several factors: coding models like Opus 4.5 / GPT-5.4 becoming sufficiently capable, matured infrastructure for AI coding agents, larger context windows up to 1 million tokens, and increased cloud GPU capacity. Other emerging trends include the mass adoption of coding harnesses by non-developers, engineers increasingly building efficient agent environments, and companies aggressively optimizing AI spend-per-token.

Key takeaway

For AI Engineers and MLOps teams deploying or managing AI agents, prioritize cloud-native architectures. Your strategy should account for agents running asynchronously in sandboxed cloud environments, leveraging increased GPU capacity and larger context windows for complex, long-running tasks. Investigate platforms like Anthropic's Claude Managed Agents or CDEs like Ona to streamline deployment, reduce local resource strain, and enable persistent agent operations, ensuring scalability and efficiency in your AI workflows.

Key insights

Cloud-based AI agents are becoming the standard for long-running tasks, driven by improved models and infrastructure.

Principles

• Cloud environments simplify agent setup and execution.
• Long-running agents benefit from cloud persistence.
• Increased context windows enable complex agent tasks.

Method

Deploy agents in sandboxed cloud environments for long-running, asynchronous tasks, leveraging large context windows and dedicated GPU infrastructure. Implement agent "confession" mechanisms for error reporting.

In practice

• Use cloud services for agent orchestration.
• Integrate agents into existing communication platforms.
• Explore CDEs for agent development and deployment.

Topics

• Cloud Agents
• AI Agent Orchestration
• Cloud Development Environments
• LLM Infrastructure
• Anthropic Claude
• OpenAI Codex

Best for: CTO, AI Architect, Investor, AI Engineer, MLOps Engineer, Director of AI/ML

Related on AIssential

• How to Use Claude Managed Agents? — Anthropic's Claude Managed Agents simplifies AI agent deployment by providing a hosted, stateful, and secure execution environment.
• OpenAI acquires AI agent orchestration startup Ona — Cloud-based sandboxes enable persistent, secure, and resource-rich execution for long-running AI agents, overcoming local machine limitations.
• Docker AI for Agent Builders: Models, Tools, and Cloud Offload — Docker provides a composable, declarative infrastructure for building and deploying complex, multi-model AI agent systems.
• Long-running Agents — Long-running agents overcome AI limitations by externalizing state, enabling sustained, multi-session progress on complex tasks.
• Anthropic's Opus 4.6 and OpenAI's Codex: New AI Models Compete — AI competition intensifies with new agent team capabilities, expanded context windows, and enterprise platforms.
• Inside Anthropic’s 2026 Developer Conference — AI platforms are evolving beyond text completion to integrated, scalable agentic systems with built-in orchestration and learning.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Pragmatic Engineer.