Forward Deployed Engineering at Cursor — Pauline Brunet
AI Engineer World's Fair 2026 · 20:47
AI coding agents and developer tools
Cursor builds AI coding tools that help developers and software teams plan, write, test, and review software. Its desktop environment and CLI support interactive development, while cloud agents work in parallel on their own computers, building and demonstrating features for review. Cursor also works in Slack, reviews GitHub pull requests, and offers scheduled or triggered automations for maintenance tasks. Developers can choose among Cursor’s own models and models from other providers.
Michael Truell, Sualeh Asif, Arvid Lunnemark, and Aman Sanger founded Anysphere, the company behind Cursor, in 2022; Truell is CEO. Its research connects model training directly to the development environment. For Composer 2, Cursor combined continued pretraining of Kimi K2.5 with reinforcement learning using the same tools and harness as the deployed agent. Its CursorBench evaluation draws on engineers’ real coding sessions, including ambiguous requests requiring changes across many files.
By August 2026, the company reported that business customers accounted for almost 80% of revenue, compared with more than 80% coming from consumers at the start of 2025. NVIDIA adoption exceeded 30,000 weekly active users within the first few months of the partnership. SpaceX completed its acquisition of Cursor in August 2026, in a reported $60 billion all-stock transaction.
AI Engineer World's Fair 2026 · 20:47
AI Engineer Code 2025 · 18:08
Affiliations reflect their AIE appearances, not necessarily current employment.
Start here to understand how continuously refreshed problems and intermediate grading signals help assess coding systems on increasingly long tasks.
Naman JainAI Engineer Code 2025
Learn how slash commands and subagents support parallel coding, Best-of-N model comparisons, and work across repositories.
David GomesAI Engineer Europe 2026
Learn how approximately six-week implementation cycles shape the recruiting and retention of engineers suited to collaborative customer work.
Pauline BrunetAI Engineer World's Fair 2026
Lee Robinson explains how Cursor built Composer, its low-latency agentic coding model, balancing coding quality with faster token generation.
Lee RobinsonAI Engineer Code 2025
Robinson connects model improvement to realistic evaluations, production-aligned training environments, and user feedback. Jain examines how evaluation changes as tasks grow from short completions to repository-scale optimization, including contamination and test manipulation.
Zakariasson describes oversight through observability, specifications, and architectural guardrails. Gomes presents a concrete tradeoff: replacing a heavyweight implementation with a Markdown skill reduces complexity but weakens isolation guarantees and discoverability.
Brunet distinguishes embedded engineering from advisory support and self-service according to customer maturity and customization needs. Her talk emphasizes clear outcomes and tightly scoped engagements rather than treating FDE as staff augmentation.
Affiliations reflect each recorded session, not necessarily current employment.