TechCapsules

Meta has launched Muse Gadgets, an open‑source initiative that lets developers attach its personal AI agent, Muse, to off‑the‑shelf hardware such as ESP32 boards, Raspberry Pi, or other Linux computers. The company released device SDKs and firmware under the Apache 2.0 license, enabling projects ranging from e‑ink morning briefings to HDMI‑connected TVs. To showcase the concept, Meta produced the Muse Home Link, a USB‑C‑powered Wi‑Fi 6 hub built on an ESP32‑C5 chip, with 5,000 units offered free to [...]

Capcom announced at its RE:2026 Open Conference that its long‑used RE Engine will be upgraded into an “AI‑generation game engine” under the REX Project. Programmer Satoshi Ishida cited three drivers: slow build times from expanding game data, the need for multilingual and accessibility support for over 2,000 developers, and the requirement to embed AI directly into the pipeline. Planned additions include RE:Dox, RE:UI, RE:Log, RE:Flows, RE:Runtime, and an optimized RE:C++ layer. The goal is to enable AI [...]

New York State draws a diverse, international student body to its numerous engineering and AI programs, offering a broad spectrum of disciplines from computer science to civil engineering. The state hosts the nation’s highest concentration of graduate AI degrees, with over 20 master’s and doctoral tracks in machine learning, computational linguistics, and intelligent systems at institutions such as New York University, Columbia University, and Cornell University. Demand for machine‑learning specialists is surging, reflected in an average New York [...]

Anthropic has reportedly embedded its response to the emerging AI tools Dots and Muse directly within Claude, the company’s flagship conversational model. The integration aims to enhance Claude’s capabilities in code generation, reasoning, and creative assistance, positioning it as a more comprehensive alternative to the specialized functionalities offered by Dots and Muse. By consolidating these features, Anthropic seeks to streamline user experience, reduce the need for multiple AI services, and reinforce Claude’s competitiveness in the rapidly evolving [...]

Google announced that, beginning October 9, the Gemini app will restrict model access for non‑subscribing users. Free users will be limited to the Flash‑Lite (3.5) model, losing access to Flash (3.6) and Pro (3.1). AI Plus subscribers ($4.99/month) will retain Flash‑Lite and Flash but also lose Pro. In contrast, AI Pro ($19.99/month) and AI Ultra ($99.99–$199.99/month) retain all existing models, and the higher‑tier plans will gain the new Deep Think capability, described as “maximum parallel reasoning.” The app will also introduce “low,” [...]

Google is tightening access to its Gemini AI models in the Gemini app. Beginning October 9, free users will be restricted to the Flash‑Lite (3.5) model, losing the ability to use Flash (3.6) and Pro (3.1). AI Plus subscribers, who pay $4.99 per month, will retain Flash‑Lite and Flash but also lose Pro. In contrast, AI Pro ($19.99/month) and AI Ultra ($99.99–$199.99/month) customers will keep all existing models and gain a new Deep Think option that enables “maximum parallel reasoning.” The update [...]

SpotEx, a cryptocurrency exchange, offers 17 assets and 25 trading pairs, and introduces a proof‑of‑reserves system that cryptographically verifies customer balances. The platform enhances community engagement through coin voting, faucets, and listings. Concurrently, Microsoft highlights the security risks of persistent AI memory, noting that memory-enabled agents can be poisoned by attackers who embed misleading data, thereby influencing future decisions. The company stresses that AI memory must be protected with the same rigor as customer data, as memory [...]

IBM has broadened its research ties with India’s Indian Institute of Science and IIT Bombay, extending collaborations first launched in 2018 and 2021. The new agenda targets quantum‑centric supercomputing workflows, hybrid Quantum‑HPC algorithms, agentic AI and sovereign Indic language models. At IISc, IBM Research India and faculty are co‑designing quantum subspace iteration methods that blend classical approximation‑tolerant diagonalization with IBM quantum processing units, aiming to boost efficiency on cloud‑native supercomputing clusters. At IIT Bombay, teams will refine [...]