Duly Noted

While experienced coders can definitely get better results via an interactive, back-and-forth code editing conversation with an agent, these results show how capable some of these models can be, even with a very short prompt on a relatively straightforward task.

Fun look at one-shoting a common task. But I love the conclusion. Experience plus iterations wins every time. I kinda like watching these as the evolution of chess robot champions…

_{Quote Citation: Ars Technica, “We asked four AI coding agents to rebuild Minesweeper—the results were explosive - Ars Technica”, 2025-12-19, https://arstechnica.com/ai/2025/12/the-ars-technica-ai-coding-agent-test-minesweeper-edition/}