Surge AI at Kaggle Bench Summit: why frontier models still struggle with work-like benchmarks
echen · x · 2026-10-28
Surge AI is speaking at the Kaggle Bench Summit in SF on what happens when benchmarks start looking less like benchmarks and more like real work: ambiguous prompts, messy environments, and the judgment calls professionals make daily.
Eric Chen recalls doing his first Kaggle competition 15 years ago and notes AI is now far better at training models than he ever was — so he's excited for Surge's benchmarks to be on leaderboards instead. The talk covers lessons from building DAYJOB, sudo L7 and GDP., and why frontier models still struggle with them.
More from Companies & People
- JHU CLSP launches student-run mentoring program for CS PhD applicants — AkariAsai · 2026-10-29
- IBM TechXchange: Three Big Challenges of Adopting AI Agents in the Enterprise — craigmullins · 2026-10-29
- 80% of Google Cloud Customers Now Run Production AI Workloads — craigmullins · 2026-10-29
- Researcher slams model release as 'marketing bullshit', says hosting Kimi K3.1 Pro would be better — eliebakouch · 2026-10-29
- AI Horizons Forum open for applications until Nov 30, aiming to onboard newcomers into AI safety — ben_j_todd · 2026-10-29
- OpenAI's Berlin DevDay: Dots coming to Europe in weeks, German Codex users up 20x — The Decoder · 2026-10-29