Ashwin Malshe

@malshe.bsky.social

833 Followers 950 Following

Associate Professor of Marketing at University of Texas at San Antonio. Visiting Professor at University of Mannheim. Machine learning, data visualization, marketing-finance interface, and R and Python coding.

Screenshot of a post from METR Evals on X.com discussing an early evaluation of “Claude Mythos Preview” for AI risk assessment. The post says the model showed a “50%-time-horizon” of at least 16 hours on METR’s task suite, with a 95% confidence interval of 8.5 to 55 hours. Below the text is a chart plotting AI model capability over time, with model release date on the x-axis and task length at 50% success rate on the y-axis (from minutes to 16 hours). Points labeled GPT-4o, Sonnet 3.7, o3, GPT-5.2 (high), Opus 4.6, and Mythos Preview trend upward along a rising green line, suggesting rapid growth in long-duration task performance. The chart notes a doubling time of 105 days and growth rate of 1020% per year.
It’s a Thor Hammer replica from Hobby Lobby priced at $9.99
A framed portrait is lying face-up inside a gray mesh office trash can lined with a clear plastic bag. The bin sits on a striped carpet next to a beige wall, and the photo appears to show part of a person in formal clothing with a U.S. flag in the background.
A vertical bar chart titled “Artificial Analysis Text to Video Leaderboard (as of November 30, 2025)” comparing ELO scores for five text-to-video models. The bars show: Runway Gen-4.5 at 1247, Veo 3 at 1226, Kling 2.5 at 1225, Veo 3.1 at 1223, and Sora 2 Pro at 1206. Runway Gen-4.5 is displayed in a bright orange textured bar, standing out against the light-gray bars for the other models. The y-axis is truncated, starting around 1200 rather than zero, which visually exaggerates the differences between the scores even though they differ by only about 20–40 points.
College completion rates are consistently higher for women across White, Black, Hispanic, and Asian ethnicities