Prediction: Mathematics will be the first science to fall to AI. This past weekend, I solved three decades-old Erdős problems using AI. That brings my total to four - all with proofs checked by computer. The models deserve enormous credit. My contribution is building the system around them and guiding their work, much as I did with ARC-AGI. The reason math will be the first science to fall is simple: computers can check mathematical proofs. AI can try an idea, discover what fails, correct it, and keep going. Each verified result becomes a foundation for the next. This loop is now accelerating, opening a path to discoveries beyond human comprehension. Physics, chemistry and other sciences have slower feedback loops. Physical experiments take time. This buys those sciences time, but AI is coming for them too. Over time, human-only science may become an art form, pursued for curiosity, beauty and meaning, while AI drives the frontier of discovery. Question now is whether human relevance also is on the menu?

Just scored 76.11% on ARC-AGI 2 — beating public GPT-5.2 and Gemini-3-Pro baselines by >20%, and (as far as I know) the best publicly reported result so far. Approach: what I’d call Multi-Model Reflective Reasoning - Using GPT-5.2, Gemini-3, Opus 4.5 - Long-horizon/multi-step reasoning (~6hrs/problem) - Agentic codegen (>100,000 python calls) - Visual reasoning - Council of judges Fun fact: all solver code was written by Gemini-3-CLI. Does this count as AI generating a new AI that beats the prior SOTA? 🤔 Full run + code (open source): kaggle.com/code/johanland/jo… @GregKamradt , holiday break is over 🙂 semi-private when? #ARCAGI #AIResearch

J Wales: "The majority of edits... are from... 600-1000 people". Waat? Not in line with #csn12 data? 10:30 @ alturl.com/ortjz