Peter Gostev
Head of AI Capabilities, Arena (LMArena) at Arena
Peter Gostev is Head of AI Capabilities at Arena (LMArena), the community platform where users vote in blind tests to rank AI models. He previously served as Head of AI at Moonpig, applying generative AI to customer service and computer vision to greeting-card discovery, and built a large following sharing hands-on tests of frontier models. Gostev began his career in strategy and analytics at Accenture before leading AI strategy at NatWest Group. He created BullshitBench, an open-source benchmark testing whether language models recognise nonsensical questions rather than confidently answering them. On Day One, he discussed how AI models are really judged.
In their own words
From How AI Models Are Really Judged, with Peter Gostev (Arena / LMArena) · In The Blink Of AIA model can pass every test you write and still produce something that looks completely awful.
Career
- Now Head of AI Capabilities, Arena (LMArena) · Arena
- Earlier · Scale-up Head of AI · Moonpig
- Earlier Led AI strategy · NatWest Group
- Earlier Strategy and analytics · Accenture
Sources: ai.engineer
Connections
Spot something wrong? Suggest a correction Updated 25 September 2026
Want a show of your own?
W2D1 Media — the team behind the Day One Network, which helped build Blackbird Ventures' Wild Hearts — builds and produces podcasts for funds, founders and operators. If you want a show built around your brand, let's talk.