Descriptions:
The AI Engineer Singapore conference Day 1 recording captures a full slate of practitioner talks from engineers and leaders at frontier AI companies, organized by Singapore’s 65 Labs builder collective and AI Engineer Global (led by Swyx, who is Singaporean). The event followed a day of 20 hands-on workshops across five concurrent tracks, with the main stage focused on production-grade AI development.
The most data-rich session comes from Mark Doyle, a software engineer at Stripe, who details “Minions” — Stripe’s internal one-shot end-to-end coding agent platform. Doyle shares that 91% of Stripe engineers now merge AI-assisted code daily, and that fully AI-generated pull requests have increased 500% over the past year — all within the constraints of a company processing nearly 2% of global GDP. He draws a sharp distinction between “co-pilot harnesses” (tools like Claude Code, Cursor, and Codex used interactively) and true one-shot agents that go from a Slack thread or prompt directly to a production pull request with no human in the loop.
Another talk from XAI covers the importance of statistical rigor when benchmarking AI-driven code changes — covering p-values, t-statistics, and sample sizing as necessary practices rather than optional refinements. Together, the sessions paint a detailed picture of how large engineering organizations are integrating agentic coding into daily workflows and where the quality and reliability challenges still lie.
📺 Source: AI Engineer · Published May 16, 2026
🏷️ Format: Livestream
