News Summaries

Spain’s Hot AI Startup Frees Up Powerful Language Model for Everyone

Multiverse Computing, a fast‑growing AI startup from Spain, just made a big splash by releasing a free, compressed version of its massive language model. The company’s new "HyperNova 60B" model, now available on Hugging Face, packs the same cleverness as the most advanced AI tools but in a much smaller, cheaper package that smaller businesses can actually run. The secret behind the shrink‑down model is a technology called CompactifAI, which borrows ideas from quantum computing to squeeze out unnecessary data while keeping performance intact. This breakthrough lets developers tap into cutting‑edge AI without the hefty cloud‑computing bills that usually come with giant models. Multiverse’s rise has been fueled by strong regional backing: the government of Aragón and the Spanish Agency for Technological Transformation helped fund a $215 million Series B round last year, and the Basque region has also pledged support, hoping the startup will become Spain’s next unicorn. Looking ahead, the team plans to open‑source more compressed models throughout 2026, aiming to democratize AI and make it accessible for a wider range of applications—from chatbots to data analysis—without breaking the bank.

Quantum Computers Get a Boost When Teetering on the Edge of Chaos, New Study Finds

A fresh study reveals that a special kind of quantum machine learning—called quantum reservoir computing (QRC)—works best when it operates right on the brink of chaos. Researchers discovered two sweet spots where QRC performance spikes: one just before the system starts to behave unpredictably (a time‑based onset of quantum chaos) and another at the precise line separating orderly, predictable behavior from chaotic, random motion. These two boundaries together form what scientists call the "edge of chaos" for many‑body quantum systems. Why does this matter? The edge of chaos is thought to be a fertile ground for complex information processing, offering enough flexibility to adapt while still retaining enough structure to stay useful. By tuning quantum devices to sit at these critical points, engineers could dramatically improve the speed and accuracy of tasks like pattern recognition, data classification, and other AI‑related functions that run on quantum hardware. The findings give a clear roadmap for building more powerful quantum technologies: aim for the narrow window where order meets disorder, and let the quantum reservoir do the heavy lifting. As quantum computers move from labs to real‑world applications, harnessing this delicate balance could be the key to unlocking their full potential.

How Digital Twins Are Turning Factories Inside Out: Real‑World Success Stories

Traditional manufacturers often feel like they’re working in a "black box" – managers can’t see what’s happening on the shop floor, equipment failures are only noticed after a shutdown, and quality problems surface only at the final inspection. Digital‑twin technology cracks open that box by creating a live, 1:1 virtual copy of the physical plant. Lanshi Group, a long‑standing equipment maker, built a digital command center that mirrors every machine and process. By feeding real‑time data into AI‑driven algorithms, the company optimized laser‑cutting routes for steel plates, boosting material utilization to about 90% – roughly 20 points above the industry norm. The twin also evolved into a predictive‑maintenance system that forecasts breakdowns and automatically issues corrective commands, earning Lanshi the Ministry of Industry and Information Technology’s “Excellent Smart Factory” badge. Meanwhile, Lianhua Auto launched the world’s first fully 3‑D‑twin plant, covering 95% of its production steps. In the virtual world, an aluminum sheet is shaped, welded and assembled in perfect sync with the real line, allowing engineers to verify and commission equipment online before the physical launch. Xingcheng Special Steel deployed a multi‑level twin across its blast‑furnace, steel‑making and rolling processes. Over 100 vertical models gave operators clear visibility, cutting abnormal furnace temperature time by 84.8%, improving hot‑blast efficiency by 12%, and lifting overall output by 11.5% while halving delivery times for heat‑treated plates. The takeaway for any factory: start with the biggest pain points – high‑energy processes, equipment that fails often, or quality‑unstable steps – and pilot a digital twin there. The results speak for themselves: greater transparency, fewer unplanned stops, higher efficiency, and a faster path to a truly smart factory.