“I think 2025 is the year that we go from ChatGPT being this super-smart thing that can answer any question you ask to ChatGPT doing things in the real world for you.”

Kevin Weil CEO

Chief Product Officer, OpenAI

Said Jan 21, 2025Deadline Dec 31, 2025On-stage interview with Axios' Ina Fried at the World Economic Forum, Davos, January 21, 2025 (reported by Axios, January 27, 2025)

OpenAI did ship agentic products in 2025 — Operator (January), Deep Research, ChatGPT agent (July) and the Atlas browser (October) — so ChatGPT gained the ability to carry out multi-step real-world tasks, but reliability and everyday adoption stayed limited, and enterprise value from agents remained the year's central point of dispute.

CapabilitiesPartially correct

“I think this is the year that, at least by the competitive coding benchmark, AI becomes better than humans at competitive coding — and stays better, forever. At the rate we're going, I would be surprised if it's 2027. I think it's going to be sooner. (o1-preview was about the millionth-best competitive programmer in the world, o1 was in the top 1,000, and o3 ranked 175th.)”

Kevin Weil CEO

Chief Product Officer, OpenAI

Said Mar 16, 2025Deadline Dec 31, 2025Overpowered podcast with Varun Mayya and Tanmay Bhat, reported by Business Today and OfficeChai, March 16, 2025

In September 2025 OpenAI's reasoning system scored a perfect 12/12 at the ICPC World Finals — enough for a first-place human ranking, ahead of all 139 university teams — and models took IOI gold, so AI matched or beat the best humans in standard algorithmic contests. But no AI held the top human competitive-programming rating outright by year-end, and a human (Psyho) beat OpenAI's model at the AtCoder World Tour Finals heuristic contest in July 2025, so the 'better forever' framing was not established.