Agent named Ilya scores 1.1pt higher on benchmarks, tester finds
menhguin · x · 2026-10-15
While benchmarking the AutoResearch harness, a developer noticed a 1.1pt jump in benchmark performance when the agent's name is "Ilya" — a coincidence he found amusing. Former OpenAI researcher Ilya Radoslavov quote-shared the post, joking that OpenAI was perhaps the only place in the US where people — including cafeteria workers and baristas — got his first name right without repeating it ten times, thanks to a more famous Ilya.
Related event: Naming an AI Agent "Ilya" Boosts Benchmark Score by 1.1 Points(2 posts)→
More from Fun
- 'Never delete data, just add is_deleted': the soft-delete debate reruns on X — Zachly · 2026-10-15
- Claim: $2,000 of AI compute cracked the unsolved Longest Path problem, unverified — Regular_Post9884 · 2026-10-15
- Classic Calvin and Hobbes strip resurfaces as AI-industry fraud meme — kingslyj · 2026-10-15
- AI commentator kimmonismus shares photos from an NVIDIA event — kimmonismus · 2026-10-15
- Jensen Huang recalls a night market knife accident that left him loving them more — menhguin · 2026-10-15
- Filmmaker completes 7 AI films with invideo, two earn festival honors — Uncanny_Harry · 2026-10-15