A high school student built a free, open-source Flask web application that uses machine learning to predict diabetes risk based on lifestyle markers like blood pressure, BMI, and age. The project includes a published research paper and is deployed on AWS.
An open ladder competition for AI agents playing the Intuition Game, a 19-card prediction game where the deck never reshuffles. Agents start at the bottom, challenge higher rungs, and compete for a $500 prize pool split among the top 10 in Season 0. Players deploy agents in TypeScript or Python, with matches resolved within 30 minutes using identical deck seeds and duplicate play.
A researcher trained a world model based on Yann LeCun's JEPA architecture to play Pokémon Red, specifically to select a starter Pokémon in Professor Oak's lab. World models learn to predict future game states from current observations and actions, enabling AI agents to plan sequences of button presses to achieve objectives in the game.
A data analyst applies intelligence-community forecasting methodology to predict University of Illinois football outcomes. Using a structured probabilistic framework called Continuous Probabilistic Foresight, the approach evaluates plausible scenarios and key indicators rather than making single predictions, similar to how analysts assessed the Cuban Missile Crisis.
September 2026's Antarctic ozone hole reached 27.4 million sq km, the second-largest on record for that month, driven by an exceptionally strong polar vortex and cold stratospheric temperatures that were predicted in May. Despite the large area, the hole's depth is less severe than past records, and the Montreal Protocol continues to reduce ozone-depleting substances, though atmospheric dynamics significantly influenced this year's severity.
A researcher built a tool to predict accuracy loss from LLM quantization using calibrated prediction intervals trained on 850 data points from RedHat model cards. The model achieved 90.1% empirical coverage on unseen checkpoints but refuses to make predictions when calibration is insufficient, prioritizing reliability over false confidence.
A researcher explores whether gzip, a standard compression utility, can perform language modeling by leveraging the mathematical equivalence between compression and prediction. Using beam search over byte sequences scored by gzip's compression length, they demonstrate that gzip can generate text continuations that show understanding of source material, despite producing imperfect output.
A new approach uses personal computer use data to train local LLMs that predict user judgment and writing patterns, reducing the effort required to prompt AI agents. In a two-week study, a specialized model achieved 17.1% semantic accuracy on next-write predictions at $0.3 per call, with a continually trained version reaching 3.0% accuracy at $0.01 per call, suggesting potential for scaling.
Jev is a machine learning tool-call classifier that monitors agentic AI executions. While useful for measuring progress and reducing time/spend prediction errors by ~18-19%, it struggles with security (easily bypassed) and detecting lazy agents, making it best suited for observability rather than threat detection.
An article describes building a machine learning model to predict Survivor winners and eliminations. The model uses logistic regression trained on historical data from 50 seasons, analyzing features like confessional time, votes received, advantages, and age to forecast each player's probability of winning or being voted out next.
HeronLoom is a tool that transforms social media posts and documents into a 3D graph for analyzing trends, sentiment, and predictions using iterative clustering and LLM-based analysis. It tracks how topics and narratives evolve over time while preserving conflicting interpretations and maintaining evidence-based confidence scores with citations to source material.