Humanity’s biggest progress takes time.

Curing cancer, going to the Mars, solving hundred-year-old mathematics puzzle… Humanity’s biggest challenges take years, decades or beyond. But today’s AI operates in short context window that lasts minutes, hours or days.

Chengxi Taylor is Co-founder & President of General Reasoning, a London-based AI research company. Her work focuses on long-horizon, non-stationary decision-making: building AI systems that can pursue a strategy over long period of time (years, decades and beyond) while adapting as the world changes.

She believes long-horizon thinking is not just a technical challenge, but a mindset. It’s about rebuilding the ability to sustain purpose and progress over time, in machines and in ourselves.

Speech

Scaling to Long Horizons

AI Engineer World’s Fair 2026, San Francisco

Research Paper

KellyBench: A Benchmark for Long-Horizon Sequential Decision Making

arXiv:2604.27865

As featured in