America today: All-in bet on AI.

Knut Jägersberg Chidambara .ML. Lionel Costes Kevin Corella N. 🔬🔭 🇪🇺 Nikos Kafritsas Mack - Tech4ll Deus ex Machina BusinessIntelligence HeidiLeeStockenstrom☮️ Mike Flache Nathalia Le Hen Eveline Ruehlin Estrella 💙 #TechForGood 💙 Kevin Corella N. 🔬🔭 🇪🇺 Fati Sule Dr. Khulood Almani | د.خلود المانع

It’s disappointing to see papers still repeat common mistakes. Each model has an optimal configuration for its specific use case. ➡️ Time series foundation models perform best with long contexts (512–2000 tokens), especially on synthetic, highly seasonal data. ➡️ Shorter

Time Series Foundation models work best with longer context-lengths. Ideally, context-length>500 Below are 2 images from TimesFM-2.5 forecasts on Kaggle-Tabular-2022 dataset: > Left: Context Length= 400 (misses the peak) > Right: Context Length = 1000 (captures the peak) Both

Testing the new TimesFM-2.5 for zero-shot forecasting. The fruits of a productive weekend!

AI is revolutionary, but certain deal structures create risky financial loops that could tank the economy. • Companies buy resources with cash. • Part of the payment comes as stock. • A major announcement drives the stock price up. • The stock surge effectively reimburses

Many people are worried about a potential AI bubble. Those who see no reason for concern often argue: "If AI drives current economic growth, why would big tech risk so much unless they were sure it would work?" But what if there was never much risk to begin with? Last week,

The success of Kimi-K2 has forced OpenAI to prematurely release GPT-5.1