The crazy story of this developer

On July 3. Consequently, 2025, in Tokyo, a certain Dog overturned from his throne one of the best AI of Openai. Furthermore, His real name: Przemysław Dębiak. Similarly, a Polish developer, who was a former employee of the firm, now converted as a self -employed. In addition, Three days of competition. Nevertheless, an almost nonexistent sleep, and a final challenge: solving an algorithmic optimization problem over ten hours in a row … Moreover, Faced with other developers, and a generative AI in full demonstration of strength. Consequently, Psyho only made a bite of the lattereven if it came out completely washed out. Consequently, Human intelligence (that of Psyho, in any case) would not yet have enough to fear AI?

The competition that makes the devs tremble …. However. However, the AI – Crazy story this developer

The Atcoder crazy story this developer new World Tour Finals is one of the most selective programming competitions in crazy story this developer the world, only the twelve best developers of the year access, after a series of fierce qualifications. In addition, The final Heuristicshe doesn’t look like any other round. Nevertheless. Nevertheless, The candidates face only one problem, but of a complexity such that no known algorithm can produce the perfect solution.

It is therefore necessary to trick. by coding what are called heuristics: approximate algorithms, capable of producing increasingly effective solutions by dint of successive revisions. The goal is therefore to design a program capable of generating A “good enough” solution to a logistical. industrial or mathematical puzzle.

Each participant can submit as many versions of his code as they wish. but must wait five minutes between each attempt. Same machine for all, free language (among those offered by Atcoder), no access to external resources.

crazy story this developer new

But this year, a new type of competitor has entered. For the first time. an artificial intelligence crazy story this developer participated in the Atcoder World Tour Finals: a model designed by Openai. called Openaiahc similar to their models like O3. A distant chatgpt cousin (model 4 or GPT-Engineer), but optimized for iterative scoring. It is not a chatbot itself but an autonomous optimization engine, cut for competition.

No conversation or prompt: only code, subject to loop to maximize a score. The objective for Openai is to demonstrate that a model can compete with the best human brains on land. hitherto reserved for algorithmic experts. No luck for the company of Sam Altman, He failed in the face of Psyho’s know-how !

Human 1 – Machine 0

At the end of the 10 am, Psyho finished with a score of 1,812 billion, 9.5 % better than OpenAi AIwhich reached “only” crazy story this developer new 1,654 billion. For a “lambda” developer. that is to say an average developer without specific experience in advanced algorithmic optimization. a crazy story this developer score in this type of heuristic competition would be extremely low, probably in thousands or tens of thousands, at best. This means the stratospheric level reached by Psyho during competition!

Psyho admitted on X that he was barely standing after the contest. when he was already coming to chain three elite competitions in three days. « Humanity has defeated (for now!). I’m completely exhausted … I’m barely alive He said in his post.

Openai preferred Insist on the podium only on defeat. « Models like O3 are already among the best 100 in the world in code. mathematics competition. but to our knowledge, it is the first time that a model has reached top 3 in a competition of this level ». Not a word crazy story this developer new for Psyho, strangely. We could have imagined at least a little wink. especially since it is an old employee. but in a way, this remains crazy story this developer coherent with the company’s strategy. Even if the AI loses, it is better toTo put forward rather than to emphasize that a human surpassing him.

• During a prestigious code competition in Tokyo, a Polish developer surpassed an OPENAI AI on an algorithmic elite problem.
• The competition opposed humans. machine over ten hours around a complex puzzle, without a perfect solution possible, targeting the best score.
• Despite the defeat of his model, Openai preferred to salute his own advances rather than recognizing human performance.

📍 To not miss any Citron press, follow us on Google News and Whatsapp.