'Humanity has prevailed (for now!)' says former OpenAI worker, admitting he's 'barely alive' after beating one in every of its AI fashions in a coding world championship struggle
Whereas it would appear to be AI fashions are quickly overtaking the endeavours of squishy human beings, it seems some hope of staying forward of our AI overlords might stay. And on this case, that hope is personified by Przemysław Dębiak, a former OpenAI worker who has narrowly crushed one of many firm’s fashions in a coding world championship.
Dębiak was competing within the AtCoder World Tour Finals 2025 Heuristic contest in Tokyo final week, pitting his coding expertise towards not solely a number of human rivals, however OpenAI’s “OpenAIAHC”, a customized simulated reasoning mannequin just like OpenAI o3 (by way of Ars Technica). The competitors concerned contestants fixing a fancy optimisation drawback over a 10-hour interval, earlier than being scored on their achievements.
Humanity has prevailed (for now!)I am fully exhausted. I figured, I had 10h of sleep within the final 3 days and I am barely alive.I will publish extra in regards to the contest once I get some relaxation. (To be clear, these are provisional outcomes, however my lead must be large enough) pic.twitter.com/fIMo0ifNCdJuly 16, 2025
The individuals had been allowed to resolve the issue utilizing any programming language accessible at AtCoder, though they had been restricted to an identical {hardware} and a compulsory five-minute wait between coding submissions.
Dębiak, competing underneath the identify “Psyho”, earned a ultimate rating of 1,812,272,588,909, inserting him prime of the leaderboard and trouncing the AI, which scored 1,654,675,725,406, incomes it a second-place end. Taking to X to have fun his achievement, Dębiak proudly introduced, “Humanity has prevailed (for now!)”, however admitted the competitors had taken its toll:
“I am fully exhausted. I figured, I had 10 [hours] of sleep within the final three days and I am barely alive.”
OpenAI, nonetheless, appears fairly proud of its mannequin’s silver-medal achievement. Talking to Ars Technica, an organization spokesperson stated:
“Fashions like o3 rank among the many prime 100 in coding/math contests, however so far as we all know, that is the primary top-three placement in a premier coding/math contest. Occasions like AtCoder give us a strategy to take a look at how effectively our fashions can motive strategically, plan over very long time horizons, and enhance options by way of trial and error—identical to a human would.”
Formally, OpenAI is perhaps right. Nevertheless, it is price noting that two Google DeepMind programs, AlphaProof and AlphaGeometry 2, earned themselves a silver-medal-equivalent rating in one of many world’s hardest maths competitions final 12 months, though neither mannequin formally entered into the competitors itself.
This specific competitors is believed to be the primary time an AI mannequin has instantly competed towards human programmers in a coding contest, and whereas I do not wish to knock Dębiak’s achievement, it is fairly spectacular that the machine got here in second place.
That in all probability does not bode effectively for the way forward for human coding rivals, as whereas programmers are inevitably enhancing their expertise, the immense pace of AI improvement is prone to imply it will not be lengthy earlier than AI fashions are topping the chief boards in related occasions.
For now, although, the champagne bottle and winner’s laurels are firmly in Dębiak’s fingers, though being a coding competitors, it was in all probability an power drink adopted by an eventual lie-down. Chalk one other one up for humanity, although—we have nonetheless obtained it the place it counts. Nearly, I reckon.
Finest gaming keyboard 2025