Gergely Orosz · 7 min read · Gen AI Tools
Impressions from visiting OpenAI, Anthropic, & Cursor
Read the original on newsletter.pragmaticengineer.com →
The author shares impressions from visits to OpenAI, Anthropic, and Cursor in San Francisco, highlighting a shift toward cloud-hosted AI agents, widespread adoption of coding harnesses by non-developers, and a pivot for engineers toward building efficient environments for agent execution while managing costs. The piece also notes practical implementations like Claude in Slack and managed-agent initiatives as indicators of how the trend may unfold industry-wide.
Key Takeaways
- Cloud agents are the next mega-trend, with hosted agent runtimes (e.g., Claude Managed Agents) gaining prominence over local setups.
- Non-developers are adopting coding harnesses at scale (OpenAI reports over 95% Codex usage among non-engineers), signaling broader reach of AI-assisted coding.
- Engineers will increasingly focus on creating efficient environments for agents to run, optimizing infrastructure rather than relying on local compute.
- Companies are aggressively optimizing spend per token, driving platform teams to reduce token costs (as illustrated by Coinbase case studies).
Excerpt
Scheduling note: this week, I’m in San Francisco at the AI Engineer’s World Fair , so there won’t be an edition of The Pulse on Thursday. However, tomorrow (Wednesday) there will be a special podcast episode – the lengthiest, most detailed one yet – with software engineering legend, Kent Beck. In recent days, I’ve visited the offices of OpenAI, Anthropic, and Cursor, in San Francisco. Onsite, I talked with software folks working on the model side to learn more about how their way of building software is changing. This article is based on observations from those visits, including some new…