9 discoveries so far. Newest first. Last updated October 27, 2026.
Computing October 26, 2026
An AI-assisted search found a serious security hole in PHP, the programming language behind a large share of the world's websites, that had been there for about ten years. After it was revealed on October 21, 2026, the U.S. National Security Agency and other agencies issued urgent warnings, and organizations worldwide rushed to fix it.
Why this is new because of AIPHP has been used and checked by countless programmers for decades, yet none of them caught this flaw in ten years. An AI combed through the code and spotted what people had missed, in time for the hole to be patched before more harm could be done.
Climate October 22, 2026
Google DeepMind used AI to build a global map that sorts every 10-meter patch of land on Earth, about 33 feet across, into distinct forest types. It goes beyond earlier maps that only showed whether trees were present.
Why this is new because of AIOlder satellite maps could tell you where trees were, but not whether a spot was old natural forest, a planted tree farm, or something else. Doing that by hand for the entire planet at this level of detail would take people an impossibly long time. AI made it practical, giving conservationists a much clearer picture of which forests are worth protecting.
Climate October 21, 2026
Researchers built an AI tool that makes better weather forecasts weeks in advance, reaching as far as a month out. The work was reported in the leading science journal Nature.
Why this is new because of AIForecasters have long struggled with the gap between next week's weather and seasonal outlooks, and forecasts two to four weeks out have been little better than guesswork. The AI learned patterns from decades of past weather that people and traditional computer models have not been able to use well. Better warning weeks ahead can help farmers, power companies and emergency planners prepare for heat waves, cold snaps and heavy rain.
Medicine August 14, 2025
MIT researchers used generative AI to design millions of molecules that had never existed, then made and tested the most promising. Two of them killed drug-resistant gonorrhea and MRSA in the lab and in mice.
Why this is new because of AIThese are not old drugs found in a library. The AI invented the molecules from scratch, and they attack bacteria in ways existing antibiotics do not. Chemists had no practical way to search that many possible molecules by hand.
Mathematics July 21, 2025
Google DeepMind's Gemini Deep Think solved five of the six problems at the world's hardest high-school math competition, writing its proofs in plain English within the same time limit as the students.
Why this is new because of AIOnly about one in twelve of the world's best young mathematicians earn gold. A computer had never reached that level by writing human-style proofs, with no special math software doing the work.
Computing May 14, 2025
Google DeepMind's AlphaEvolve found a way to multiply four-by-four grids of complex numbers in 48 steps instead of 49, improving a method from 1969. It also found savings that recovered about 0.7 percent of Google's worldwide computing power.
Why this is new because of AIMathematicians had tried to beat this 1969 result for more than half a century and failed. The AI wrote and tested its own new algorithms until it found one people never had.
Biology October 9, 2024
The 2024 Nobel Prize in Chemistry went to the creators of AlphaFold, which predicted the 3D shape of nearly every known protein, and to David Baker, whose AI tools design brand-new proteins.
Why this is new because of AIWorking out one protein's shape used to take scientists years. AI did it for about 200 million proteins, and it now designs proteins nature never made. The Nobel committee itself said this was a breakthrough that had not been possible before.
Materials November 29, 2023
Google DeepMind's GNoME predicted 2.2 million new crystal structures, including 380,000 that are stable enough to make. Labs have already made hundreds of them.
Why this is new because of AIIn all of history, people had found about 48,000 stable crystals. The AI multiplied that roughly tenfold in one project, a pace DeepMind compared to about 800 years of normal discovery.
Medicine February 20, 2020
An MIT AI model screened more than 100 million molecules and picked out halicin, a compound that kills many drug-resistant bacteria, including some that resist every known antibiotic.
Why this is new because of AIHalicin had been studied as a diabetes drug and nobody suspected it could kill bacteria. The AI spotted it by learning patterns no human chemist was looking for.