Toggle light / dark theme

Beyond Agentic AI: The Emergence Of Cognitive AI Ecosystems

In the next decade, AI will likely undergo more significant changes than only becoming more independent; it will also grow more cognitive. AI systems will act as interconnected ecosystems that are capable of contextual awareness, cooperative reasoning, ongoing learning, and adaptive decision-making in almost every facet of society, rather than isolated applications.

Large language models of today are remarkable due to their ability to produce and anticipate information. Persistent memory, multimodal perception, long-term planning, causal reasoning, and self-directed learning within strictly regulated bounds will probably be characteristics of the AI of 2036. Similar to biological neural networks, millions of specialized AI agents will work together to create dynamic intelligence fabrics that continuously optimize national defense, manufacturing, transportation, financial markets, healthcare delivery, and energy grids.

The line between workforce and software will become increasingly hazy. Hundreds of thousands of AI agents working continuously alongside human employees may be employed by organizations as digital workforces. A customized constellation of AI advisers, researchers, legal assistants, financial analysts, engineers, and cybersecurity specialists working around the clock could be present for every knowledge worker. This shift signifies the emergence of an entirely new digital labor force in addition to automation.

Hybrid AI model cuts financial forecasting error across stocks and crypto

A hybrid artificial intelligence model that combines two well-established deep learning techniques has improved the accuracy of financial market forecasts across major stock indices and so-called cryptocurrency, according to work in the International Journal of Reasoning-based Intelligent Systems.

The researchers designed the model, CLSTM-HN, to address a long-standing problem in financial forecasting: balancing the detection of short-term market movements with the recognition of longer-term trends. The researchers tested the system on publicly available data and achieved a forecasting error 15% to 20% lower than that of conventional long-short-term memory (LSTM) models. They also saw an improvement in the accuracy of predicting whether prices would rise or fall by 10% to 14%.

Financial markets are difficult to predict because prices are volatile, noisy and subject to sudden structural shifts. Traditional statistical approaches often rely on assumptions about market behavior that break down during periods of instability.

Fake Paysafe, Skrill SDKs on NPM and PyPi steal credentials

Malicious packages on the Node Package Manager (npm) and the Python Package Index (PyPI) delivered stealer malware to developers and users of Paysafe, Skrill, and Neteller payment applications.

The threat actor published at least 17 malicious packages simultaneously, each tasked to exfiltrate credentials and access tokens to a command-and-control server hosted on Amazon Web Services (AWS).

All three payment platforms are popular, with Paysafe being mostly used by e-commerce sites and online marketplaces, gaming platforms, travel businesses, and financial services or software-as-a-service (SaaS) providers.

Hy3 (free) — API Pricing & Providers

Hy3 is a 295B-parameter Mixture-of-Experts model from Tencent (21B active, 192 experts with top-8 routing) built for reasoning, agentic workflows, and real-world production use. It supports a configurable reasoning effort: a direct no-think mode by default, plus low and high chain-of-thought modes for complex math, coding, and multi-step problems. With a 256K context window, Hy3 targets long-horizon tasks, including improved coreference resolution, multi-turn constraint tracking, and stable tool-calling that generalizes across agent scaffoldings.

Tencent positions it as a reliable, cost-effective option across coding, document processing, financial analysis, game development, and frontend design, with a strong emphasis on grounded, anti-hallucination behavior that answers when grounded and flags when evidence is missing rather than fabricating.

The world’s first ultra-compact semiconductor chip for biosignal measurement

A research team led by Prof. Junghyup Lee of the Department of Electrical Engineering and Computer Science at DGIST has become the first in the world to develop a “time-interleaved noise-shaping SAR ADC (analog-to-digital converter)” semiconductor chip capable of simultaneously measuring multiple biosignals, including electrocardiograms (ECG) and electromyograms (EMG). The team developed this technology in an actual semiconductor chip and successfully completed functional validation. Their findings were presented at the IEEE Symposium on VLSI Technology & Circuits (VLSI 2026), held in Honolulu, June 14–18.

Accurately measuring multiple biosignals using wearable devices such as smartwatches requires meeting several demanding conditions. These include “ultra-high input impedance (resistance)” to prevent signal loss even when no sweat is present on the skin or when contact is loose (dry or non-contact electrodes), a “wide input range” to prevent signal distortion caused by vigorous movement, and “ultra-low power consumption” for long-term operation. However, conventional measurement approaches have struggled to satisfy all these requirements simultaneously within a single chip.

Lee’s research team addressed this challenge by proposing a novel “time-interleaved third-order noise-shaping SAR ADC” architecture in which circuit blocks that consume significant power and chip area are shared across multiple channels, while only essential components (the residual capacitor banks) are allocated separately to each channel. This approach dramatically reduced the circuit area and power consumption required for multichannel systems, enabling an ultra-compact, ultra-low-power chip.

Strange Things Are Happening in Quantum Computing

Sign up on Freecash using my link: https://bit.ly/4eDWh6q to get $10 bonus for your first offer!

Quantum computing is supposed to be one of the most exciting new technologies that humanity is working on, with companies promising it can be used in chemistry, material science, logistics, and finance. Over the years, those use cases have been slowly eroded, but investment in quantum tech has only increased. Why? Let’s take a look.

👕T-shirts, mugs, posters and more: ➜ https://sabines-store.dashery.com/
💌 Support me on Donorbox ➜ https://donorbox.org/swtg.
👉 Transcript with links to references on Patreon ➜ / sabine.
📝 Transcripts and written news on Substack ➜ https://sciencewtg.substack.com/
📩 Free weekly science newsletter ➜ https://sabinehossenfelder.com/newsle…
👂 Audio only podcast ➜ https://open.spotify.com/show/0MkNfXl…
🔗 Join this channel to get access to perks ➜
/ @sabinehossenfelder.
📚 Buy my book ➜ https://amzn.to/3HSAWJW

#science #sciencenews #quantum

Ousaban Banking Trojan Targets Iberian Bank Users with Fake PDF Lures

The current version moves that screening to the operator’s server, so the exact rules are hidden. Either way, visitors outside Spain or Portugal get a Spanish “access denied” notice instead of malware.

Clear the check, and the download starts. A script downloads an image that looks like a PDF icon but hides a ZIP file inside, a trick called steganography. The script unpacks Ousaban from that ZIP, runs it, then deletes the image, the ZIP, and itself to leave less behind. Once running, Ousaban adds a registry entry named Financeiro (Portuguese for “finance”) so it starts up with Windows.

Ousaban’s command server, the machine that controls it, is deliberately hard to find. It carries a Pastebin link that points to one server address, but Fortinet says that address is a decoy.

Two South Korean companies named Samsung Electronics and SK Hynix now manufacture roughly two-thirds of the memory chips inside almost every digital device on Earth — produced inside a country whose 1953 per-capita income was lower than Somalia’s or Haiti’s

Open any device built in the past five years, look inside its memory subsystem, and the chips you find were almost certainly fabricated in one of three South Korean industrial cities — Hwaseong, Pyeongtaek, or Icheon — by one of two companies whose combined market capitalisation now exceeds $700 billion. The historical improbability of this situation is not a matter of degree but of category. Korea in 1953 did not have a semiconductor industry, a precision manufacturing tradition, an advanced engineering workforce, or the kind of capital markets that could finance industrial development. It had a per-capita income lower than essentially every other country whose subsequent economic trajectory has been studied by development economists, a primarily agricultural economy substantially destroyed by three years of active warfare, and a small population (~20 million) whose adult literacy rate stood at approximately 20 percent. The proposition that, 72 years later, two companies headquartered in the same country would manufacture the memory chips inside Apple’s iPhones, Google’s Pixel devices, Microsoft’s data centres, Nvidia’s AI accelerators, Tesla’s autonomous-driving computers, and essentially every other major piece of digital hardware sold globally — would have been considered, by any reasonable observer in 1953, structurally impossible.