AI Achieves Rapid Software Replication and Major Mathematical Breakthroughs
Recent AI breakthroughs feature rapid commercial software replication using clean-room AI agents and OpenAI solving complex mathematical problems, sparking intense debate among mathematicians.
The past few weeks have marked a dual turning point in artificial intelligence capabilities: the rapid replication of major commercial software alongside breakthroughs in solving complex mathematical problems.
Developer Brandon Thomas sparked international controversy when he announced in a social media post that software development as we know it is over. He described how he recreated Adobe Photoshop—estimated to include around ten million lines of code—in just three hours, and over a single weekend released open-source versions of seven major Adobe applications, including Illustrator, Premiere, After Effects, and Acrobat.
The Clean Room Approach and Reverse Engineering
According to Thomas, the motivation behind the move was retaliation against the company's cancellation fee policies, which led to a lawsuit by the US Federal Trade Commission and a $150 million settlement. Thomas performed the duplication using an AI agent in a clean room architecture, where the agent examines functionality and interface alone and writes completely new code without exposure to the original source code.
This phenomenon is not limited to him: reverse engineering projects are gaining massive momentum, led by the REA project (Reverse Engineer Anything). The project, which jumped to the first place in GitHub rankings with about 300,000 stars, allows developers to disassemble existing software and generate independent alternatives at the click of a button.
The Mathspocalypse and Advanced Mathematics
Simultaneously, OpenAI published a document detailing solutions to approximately 722 open mathematical problems, an event dubbed the mathspocalypse. Among the prominent achievements was tangible progress toward proving the Riemann Hypothesis—one of the central millennium problems in mathematics since the 19th century concerning the distribution of prime numbers—where the model managed to narrow the range of possible solutions.
Another dramatic advancement was demonstrated in Strassen's hypothesis for matrix multiplication, a fundamental algorithm used in many modern systems such as recommendation engines and graphic processing. While human mathematicians took four decades of research to reduce the efficiency coefficient from 2.38 to 2.371, the AI model lowered the metric to 2.25 within hours.
Impact on Human Skills and Education
These developments have triggered fierce reactions among senior mathematicians. Prof. Hugo Duminil-Copin, a Fields Medalist, shared that he cried at night because AI decimated his research field, calling it an attempt to dismantle their scientific activity. Prof. Terence Tao strongly criticized the use of internal, closed models to solve problems without scientific transparency, calling on the community to fight for research ethics and standards.
The growing reliance on artificial intelligence raises heavy educational, cognitive, and emotional questions. Studies point to a decline in academic achievements in sciences, reading, and mathematics, as well as an erosion in cognitive metrics among high school students. A study from MIT found that regular use of AI-based tools may lead to cognitive atrophy: independent writing quality declines, and users struggle to initiate thinking and formulation processes without technological assistance.
